English Version of the Follow-Up Sur- Explain the Purpose of the Credit with Educa- Vey Is Attached As Annex A

F REEDOM FROM HUNGER

RESEARCH PAPER NO. 4

Impact of Credit with Education on Mothers and Their Young Children’s Nutrition: Lower Pra Rural Bank Credit with Education Program in Ghana

Barbara MkNelly and Christopher Dunford in collaboration with the Program in International Nutrition, University of California, Davis

March 1998 Contents Executive Summary ...... 1 1.0 Introduction ...... 5 2.0 Impact Evaluation Design and Methods ...... 9 3.0 Survey Results: Characteristics of the Respondent Sample Groups ...... 14 4.0 Intermediate Benefits: Women’s Economic Capacity ...... 17 5.0 Intermediate Benefits: Women’s Health/Nutrition Freedom from Knowledge and Practice ...... 31 Hunger’s Mission 6.0 Intermediate Benefits: Women’s Empowerment ...... 45 Founded in 1946, Freedom 7.0 Ultimate Impact: Nutritional from Hunger promotes “Self- Status and Food Security ...... 54 Help for a Hungry World.” Freedom from Hunger brings Résumé Analytique ...... 60 innovative and sustainable Resumen Executivo ...... 65 self-help solutions to the fight Bibliography ...... 69 against chronic hunger and poverty. Together with local Index of Figures and Tables ...... 70 partners, we equip families Appendix ...... 72 with resources they need to build futures of health, hope This evaluation research was conducted and dignity. with the support of the Thrasher Research Fund Innovation Grant Award #02902-5. Additional support was received from the Nutrition Division of UNICEF/New York and PLAN Credits International Design ©1998 Freedom from Hunger Michael Curry/ No part of this document may be Donna Justice reproduced without the express written permission of Freedom from Hunger.

Credit with Education is a service mark of Freedom from Hunger.

1644 DAVINCI COURT DAVIS, CA 95617 (530) 758-6200 FAX (530) 758-6241 E-MAIL: [email protected] IMPACT OF CREDIT WITH EDUCATION ON MOTHERS AND THEIR YOUNG CHILDREN’S NUTRITION: LOWER PRA RURAL BANK CREDIT WITH EDUCATION PROGRAM IN GHANA

BARBARA MKNELLY AND CHRISTOPHER DUNFORD IN COLLABORATION WITH THE P ROGRAM IN I NTERNATIONAL NUTRITION, UNIVERSITY OF CALIFORNIA, DAVIS

EXECUTIVE SUMMARY 1993 baseline and a 1996 follow-up—with different mother/child pairs participating in Since 1989 Freedom from Hunger has the two time periods. A quasi-experimental worked with local partners to develop and design was applied at the community level disseminate a cost-effective integrated pro- to minimize possible bias. Following baseline 1 gram strategy called Credit with Education, data collection, 19 study communities were with the goal of improving the nutritional assigned to either a “program” or “control” status and food security of poor households group, with the latter not to receive Credit in rural areas of Africa, Latin America and with Education until after completion of the Asia. In collaboration with the Program in evaluation research. International Nutrition at the University of California, Davis, Freedom from Hunger un- Three sample groups of women with children dertook a multi-year study in Credit with Edu- under three years of age were included in the cation program sites in Bolivia and Ghana. follow-up research: (1) Credit with Education Financial support for this collaborative re- program participants of at least one year, (2) search was provided by an Innovations nonparticipants in program communities Grant from the Thrasher Research Fund, and (3) residents in control communities se- with supplemental funding from the Nutri- lected not to receive the program for the pe- tion Division of UNICEF/New York. riod of the study. Women for the two nonparticipant groups were randomly selected The evaluation research was designed to test from comprehensive lists of all women with hypotheses of positive program impact on children under three years of age. children’s nutritional status, on their mothers’ economic capacity, women’s empower- Program impact is evaluated by comparing ment and mothers’ adoption of key child sur- the differences between the responses and vival health/nutrition practices. measurements in the two data collection rounds (1993 and 1996) for program par- This report presents the results from the ticipants versus nonparticipants in program evaluation research conducted in the Lower communities and residents in control com- Pra Rural Bank Credit with Education program munities. Different sets of women were in- area in coastal Ghana. Two major survey cluded in the two data collection rounds, be- and anthropometric (heights and weights) cause few women had under-three-year-old data collection rounds were carried out—a children in both 1993 and 1996. Baseline 1Credit with Education is a service mark protected by Freedom from Hunger for the exclusive use of member organizations of the Credit with Education Learning Exchange.

Research Paper No. 4!1 respondents in the program communities had net monthly enterprise incomes as high were later reclassified on the basis of whether as $200 to $300 per month, but 10% had they ever joined the program when it was ul- net incomes of $10 or less. Diversity of in- timately offered in their community. Conse- come impact was clear even within the same quently, baseline respondents in the study Credit Associations, with some women en- communities that received the program are joying considerable improvement in their eco- classified either as “future participants” or nomic activities and others experiencing little “future nonparticipants.” By comparing the change. A better understanding of the fac- 1993 baseline measures of “future partici- tors that allow some women to be relatively pants” to actual participants in 1996, the more successful—individual attributes, en- difference between years can better be attrib- trepreneurial skill, specific loan activities or uted to the impact of the program and not program loan terms—could stimulate to inherent differences between women who changes in program implementation, such as self-select to join the Credit with Education pro- incorporating basic business management gram and those who do not. education, which might enhance the eco- There was no statistically significant differ- nomic impacts for less successful borrowers. ence in the socioeconomic status of house- There was evidence that the program was holds (as measured by consumer assets) or fostering the entrepreneurial skills of partici- women’s education and literacy across the pants. Between years, participants were sig- three sample groups in either of the time pe- nificantly more likely to consider demand riods. However, participants in the follow- and profitability when deciding upon in- up period were significantly older, had more come-generating activities. There was also a children and were more likely to have re- significant difference in the percentage hav- cently engaged in a nonfarm enterprise than ing savings and the value of cash savings be- nonparticipants or residents in control com- tween years for participants versus controls munities. and participants versus nonparticipants. Impact on Women’s Economic Although nonfarm incomes had increased, Capacity there was no significantly different change in participants’ assessment of their relative The vast majority of 1996 participants contribution to their households’ total in- (90%) felt that their incomes had “increased” come as compared to the two nonparticipant or “increased greatly” since they had joined groups. There were also few significant dif- the Credit with Education program. Most com- ferences across the groups in change of house- monly, participants attributed this improve- hold expenditures on food, clothing, medi- ment to expansion of their businesses, re- cine, school expenses, house repair or busi- duced input costs as a result of buying in ness assets. It is possible that substitution bulk or with cash, and new activities or prod- of responsibility for food purchases within ucts made possible by access to credit. There participant households might be undermin- was a significantly greater increase between ing program impact on per capita food ex- years for participants’ monthly nonfarm penditures. earnings as compared to nonparticipants and residents in control communities. The in- Impact on Mothers’ Health/Nutrition crease in net nonfarm monthly income (rev- Practices enue minus costs) was $36 for participants, $18 for nonparticipants and $17 for resi- Among women who had more than one child, dents in control communities. participants in 1996 were significantly more While the 1996 participants overall exhib- likely to report positive change in how they ited significantly greater improvement in their breastfed or fed their younger children in- nonfarm earnings, there was considerable cluded in the study than were nonpartici- range in monthly profits. Some participants pants or residents in control communities.

Impact of Credit with Education !2 Relative to nonparticipants and/or residents The children of participants also experienced in control communities, participants re- significantly greater improvement in feeding ported significantly greater positive change frequency as compared to the children of the in a variety of the health/nutrition practices two nonparticipant groups, with a margin- promoted by the Credit with Education pro- ally significant difference in egg and meat/ gram: fish consumption. ! Giving newborns the antibody-rich first Impact on Women’s Empowerment milk (colostrum). Indicators of women’s empowerment were de- ! Introducing liquids and first foods (in veloped to evaluate program impact on addition to breastmilk) closer to the ideal women’s self-confidence and vision of the age of about six months. future, their status and bargaining power ! Not using feeding bottles. within the household, and their status and ! Enriching the traditional complementary networks in the community. food, koko, with bean/cowpea, egg, fish, Compared to the two nonparticipant groups, groundnut, milk and palm oil. the 1996 participants rated themselves as ! Enriching Weanimix (a complementary being significantly more confident that they food promoted and distributed by the would be able to Ministry of Health) with fish powder. ! feed their child the good foods that they ! Rehydrating children who had diarrhea know they need; by giving them either ORS (made from ! prevent their child from getting diarrhea the packets) or home liquids (like tea or and other illnesses; and rice water). ! earn more money next year than this ! Not “treating” children who had diarrhea year. by giving them enemas. However, they were not more confident that ! Knowing ways to prevent diarrhea, such they could educate their children to their as “covering food to avoid flies” and children’s full potential. “keeping food clean.” At the level of the household, participants’ Despite involvement in their loan-financed bargaining power did not significantly im- activities, participants did not wean their prove as compared to the other two groups children any earlier than did nonparticipants in decisions regarding a number of household and were just as likely to breastfeed their investments, such as how much to spend on babies into the child’s second year of life. clothing, medicine, agricultural inputs or fix- However, no statistically significant differ- ing the house. However, there was a signifi- ence was found in the following areas, indi- cant increase in participants’ “say” in cating a need for greater nonformal educa- whether or not children went to school as tion in these topic areas: compared to nonparticipants, and a marginally significant difference as compared to resi- ! Other diarrhea prevention practices pro- dents in control communities. moted by the program, such as hand washing, reheating cooked food before Participants’ husbands were significantly serving/not keeping cooked food long be- more likely to have offered to help their wives fore serving, and using clean water. with child care and with their income-generating activities during the previous six ! Limiting or withholding food from chil- months as compared to nonparticipants’ hus- dren with diarrhea, as reported by the bands; however, there was no significant dif- majority of women in each of the three ference between participants and residents groups. in control communities. There was also no ! Immunization coverage. significant difference across the groups in

Research Paper No. 4!3 women reporting they had discussed family (HAZ) for participants’ one-year-olds was planning with their husbands. almost 0.3 greater than the baseline HAZ of future participants’ one-year-old children. At the level of the community, the program The mean HAZ for children in control com- seemed to have positively affected women’s munities was 0.2 less for the same period of participation in the community and their time. A similar positive effect was not found helping contacts with family and friends. for maternal nutritional status as measured There were significantly greater changes be- by women’s body mass index (BMI). tween the years for participants as compared to the two nonparticipant groups, in that Conclusions participants were more likely to The impact evaluation research in Ghana ! be members of a community group provides evidence that credit and education beyond their families; services, when provided together to groups ! have helped a friend with his/her work; of women, can increase income and savings, improve health/nutrition knowledge and ! have offered health/nutrition advice to practice, empower women, and ultimately others; and improve household food security and ! have offered business advice to others. children’s nutritional status. Further analysis is planned to examine the relationship be- Participants were also contributing more tween the various intermediate impacts and money to non-kin funerals, which is impor- their relative contribution to children’s bet- tant to an individual’s social status and to ter nutritional status. the reputation of one’s family. Although not a focus of the impact research, Using these three aspects of women’s empow- it is also important to note the program’s per- erment, participants were significantly more formance in terms of financial sustainability. “empowered” than the two nonparticipant In the six-month period from October 1996 groups, especially at the individual and com- through March 1997, the program had an munity levels. However, it is interesting to operating self-sufficiency ratio of 81%, mean- note that residents in control communities ing that the interest paid by borrowers cov- were more confident and enjoyed relatively ered 81% of the Lower Pra Rural Bank’s costs greater assistance from their husbands than of delivering the credit and education, cov- nonparticipants in program communities. It ering all operating costs including financial is possible that the decision of nonpartici- costs such as interest on debt, but not loan pants not to join the program in their com- loss reserve. While not yet fully financially munity itself reflects an initial lack of self- sustainable, this represents a much higher confidence and greater degree of inequity in level of cost recovery than most income-gen- marital relations. eration interventions and certainly more than traditional health/nutrition education Impact on the Ultimate Goals— programs. The combination of positive im- Nutritional Status and Food Security pact and financial sustainability makes Credit with Education a strategy with exciting poten- Participant households reported a reduced tial for widespread and sustainable impact vulnerability to the “hungry season” relative on nutrition and food security. to the baseline period as compared to the two nonparticipant groups. The nutritional status of participants’ one-year-old children—both in terms of weight-for-age and height-for-age—was also significantly improved between the years relative to the children of residents of control communities. For example, the mean height-for-age z-score

Impact of Credit with Education !4 ! Program participation will increase women’s status and self-confidence to 1.0 INTRODUCTION plan and offer a healthy diet to their families, especially to their young Freedom from Hunger, in collaboration with children. the Program in International Nutrition at the University of California, Davis, undertook a The conceptual framework guiding this im- three-year impact evaluation of Credit with pact evaluation is depicted in the hypoth- Education as implemented by the Lower Pra esized benefit process diagrammed in Figure Rural Bank in coastal Ghana. Funding was 1.1. The strategy’s ultimate goals—improved provided primarily by an Innovations Grant household food security and nutritional sta- from the Thrasher Research Fund and a tus—require first that the intermediate ben- smaller grant from the Nutrition Division of efits of poverty alleviation, empowerment UNICEF/New York. and behavior change be achieved. For this reason, qualitative and quantitative meth- The evaluation research was designed to test ods were used in addition to measures of four hypotheses: nutritional status (maternal and child heights ! Credit with Education in a community has and weights) to investigate impact on the a positive effect on the nutritional status program’s intermediate goals—women’s eco- of children. nomic resources, their health/nutrition knowledge and practice, and women’s em- ! Program participation will increase powerment as measured by their self-confi- women’s economic capacity (income, dence and status. savings, time) to adopt beneficial behaviors and to invest in nutritionally As indicated on the left side of Figure 1.1, important expenditures such as food and the Credit with Education strategy has program health care. performance as well as impact goals. It is important to appreciate that the desired im- ! Program participation will increase pacts are not being pursued at any financial women’s knowledge, trial and adoption cost. Rather, the strategy is designed and of beneficial breastfeeding, weaning and implemented so that the credit and educa- diarrhea management and prevention tion services are sufficiently cost-effective to practices. allow for expansion and financial sustainability.

Figure 1.1 Freedom from Hunger Credit with Education Strategy

High-Performance Program (HPP) Characteristics Benefit Process

Large Scale Income and Credit Savings

Cost- Improved Household Effectiveness Women's Self-Confidence Food Security Associations and Status Better Health and Nutrition Financial Education Knowledge Sustainability and Practice

Program Intermediate Longer-Term Inputs Benefits Outcomes Self-Reliant Local Institution

Research Paper No. 4!5 Background on Credit with Education women, and earned in steady and regular amounts. Founded in 1946, Freedom from Hunger is an international nonprofit organization ! Income increases alone are unlikely to working to empower the poorest families and have substantial impact on the communities to help themselves overcome malnutrition of women and young hunger and malnutrition. Since 1989, Free- children unless key maternal and child dom from Hunger has developed and dissemi- health/nutrition behaviors are also nated a cost-effective and sustainable pro- adopted. gram strategy called Credit with Education to ! The scope and scale of the problems of improve the nutritional status and food se- hunger and malnutrition require curity of women in poor, rural areas of Af- solutions with potential for widespread rica, Latin America and Asia. Freedom from expansion and financial sustainability. Hunger provides training and other technical assistance to local organizations (prima- In terms of program sustainability, a high de- rily local financial institutions but also non- gree of loan recovery (as of March 1997, the profit organizations) which directly imple- repayment rate across all Credit with Educa- ment the Credit with Education programs. tion programs was 99%) and the use of real interest rates and savings build a loan fund Credit with Education combines small-scale that can be recycled again and again. Inter- loans (less than $300) with education in the est and fee payments are used to pay ad- basics of health, nutrition, birth timing and ministrative costs of program delivery, with spacing, and small business skills. Partici- full recovery of operating costs expected pants form self-managed Credit Associations within three to five years of start-up in most (village banks) and guarantee each other’s areas. Sustainability is also attained loans. The women invest their loans in in- through building or developing local capac- come-generating activities in which they are ity to implement, manage and expand pro- already skilled, then meet weekly to repay gram operations. the principal and interest and to deposit savings. Learning sessions (adapted to local Still, despite the popularity of microcredit needs) are also delivered at each meeting to and the intuitive potential of Credit with Edu- provide important knowledge on basic cation, there has been little evidence to date health and nutrition practices, family plan- of the impact of such programs on food se- ning and small business management. curity or malnutrition (Berger and Buvinic, 1989; MkNelly and Dunford, 1996; The design of Credit with Education was based Sebstad and Chen, 1996). For this reason, on “development breakthroughs” such as the Freedom from Hunger, in collaboration with Grameen Bank, evidence in the literature, the Program in International Nutrition at and Freedom from Hunger’s own experience the University of California, Davis, under- of key programmatic features that offer the took a three-year impact evaluation of Credit greatest potential to alleviate hunger and with Education in two program sites—coastal malnutrition. Some of the major assump- Ghana and the Altiplano in Bolivia. This tions underlying the design of the strategy report summarizes the findings from the include: Ghana survey. ! Inadequate access to more and better food rather than food scarcity per se is Background on the Lower Pra Rural the chief problem faced by the majority Bank’s Credit with Education Program of food-insecure households. In 1992, Freedom from Hunger and the ! Income increases that will have the most Lower Pra Rural Bank embarked upon a part- direct, positive impact on food security nership to provide Credit with Education ser- and nutrition are those earned by the vices to poor, rural women in Shama Ahanta poorest households, controlled by East District of the Western Region. This was

Impact of Credit with Education !6 the first Rural Bank in Ghana with which Free- The Credit Component dom from Hunger partnered. Currently, there As of March 1997, the Lower Pra Rural are five Rural Banks implementing Credit with Bank had made over 9,000 individual loans Education with a total membership of approximately 6,000 borrowers (see map, Figure with a total value of just under $600,000 to women participating in the Credit with Edu- 1.2). cation program since its inception in 1992. Rural Banks are autonomous, community- Although liquidity constraints have greatly hin- based organizations regulated by the Bank dered the expansion of the program to include of Ghana. They were new borrowers, the originally set up as com- Figure 1.2 Credit with Education loan repayment per- munity-managed devel- Program Areas Ghana, West Africa formance of the exist- opment banks to mobi- ing clientele has been Builsa Community Bank lize rural savings, fur- Nandom Upper East excellent—never fall- nish credit to rural en- Rural Bank Upper ing below 92% for any trepreneurs and sup- West quarter over the life port community initia- of the program. tives. Rural Banks are Northern As of March 1997, committed, by man- there were 1,491 date, to serve small- women organized in scale economic enter- 55 Credit Associa- prises in their service ar- Nsoatreman Brong-Ahafo

Rural Bank Sunyani tions in the commu-

eas, although few of V o nities surrounding their clients are women lta the Lower Pra Rural and most loans are Ashanti Bank. While the ma- much larger than the Eastern jority (1,131) of the less-than-$300 charac- Accra Greater women were taking teristic of poverty lend- Western Central Accra loans, approximately ing programs such as (capitol) one quarter (360) of Credit with Education. Brakwa-Breman Rural Bank the women were par- However, the purpose, Lower Pra ticipating in the edu- philosophy and organi- Rural Bank cation sessions and zation of the Rural Rural Bank depositing savings Banks fit well with Free- Private Savings & Loan only. The total dom from Hunger’s amount of loans out- goals and strategies. standing to these Freedom from Hunger provides training and women was $88,173, and their savings on de- technical assistance to the boards and staffs posit with the Rural Bank was $10,471. The of participating Rural Banks to implement average loan size was the cedis equivalent of the program. Freedom from Hunger’s $78 for a four-month period. Table 1.3 sum- Technical Support Center in Accra assists the marizes the most common loan activities re- Rural Banks in the organization and man- ported by borrowers taking loans during the agement of Credit Associations and in the first quarter of 1997. establishment of a system to manage the de- The Education Component livery of financial services, health/nutrition education and microenterprise education. The education component of Credit with Edu- The Rural Banks are responsible for housing cation is designed to complement the credit and staffing the program. They must pro- component by empowering women with the vide a significant portion of the loan capital information, skills and confidence they need required, and they must apply all interest to better manage their own and their fami- earned on program loans to program costs. lies’ health and nutrition. The Credit Asso- ciations’ regular meetings include learning

Research Paper No. 4!7 Table 1.3: Loan Activities Reported by Borrowers Beginning a Loan Cycle January - March 1997 Loan Activities Borrowers Make/Sell Oils (coconut, palm oil) 36% Buy/Sell Fish (fresh or smoked) 22% Make/Sell Cooked Food (kenkey, banku, chop bar) 11% Buy/Sell Non-Food Items (utensils, charcoal, clothing, 8% cosmetics, jewelry) Buy/Sell Maize 8% Tabletop Store (milk, sugar, toffees, etc.) 6% Buy/Sell Fruit (oranges, bananas, blackberries) 5% Make/Sell Local Gin (akpeteshie) 4% sessions addressing three areas: health and propriate solutions for themselves and their nutrition, microenterprise development and families. Credit Association management. The same Ideally, the credit and education components field agents that assist with the loan process reinforce each other by addressing both the facilitate these learning sessions. Field agents informational and economic obstacles to receive training in nonformal education tech- better health and nutrition. The education niques, as well as lesson plans and a curricu- promotes nutritionally beneficial spending lum for sequencing the following topics: and intrahousehold distributions, as Health and Nutrition Topics women’s increased income and productivity • diarrhea management and prevention help to overcome economic barriers to the • breastfeeding adoption of better health/nutrition practices. • infant and child feeding The success of income-generating activities • immunization financed by the program and the participa- • family planning tory program design foster change in women’s self-confidence and learning readi- Microenterprise Topics ness to adopt important practices. • choosing an appropriate activity • increasing profits The purpose of the combined services is to • increasing sales allow and encourage women to: (a) earn and • managing a microenterprise use income to gain access to adequate quality and quantity of food; (b) exclusively Credit Association Management breastfeed their infants for the first six • group formation months, if possible, and to introduce nutri- • loan analysis ent-dense complementary foods at about six • setting and enforcing rules months of age; (c) rehydrate children dur- • setting and assessing goals ing diarrheal episodes and practice personal Within the Health and Nutrition and and food hygiene to help prevent diarrhea; Microenterprise topic areas, specific ideal be- (d) seek the full immunization series recom- haviors are promoted. The learning sessions mended for infants and women, where lo- include skits, stories and demonstrations so cally available; and (e) make more informed that these topics and ideal behaviors are ad- reproductive decisions for themselves and dressed in a participatory rather than lec- their families. ture format. Lessons are also sequenced and field agents trained so that within each topic area participants are encouraged to identify problems, then analyze, test and adopt ap-

Impact of Credit with Education !8 Financial Performance A recent programmatic review undertaken by Price Waterhouse for UNICEF/Ghana as- 2.0 IMPACT EVALUATION DESIGN sessed the profitability of the Credit with Edu- AND METHODS cation program for the Rural Banks implementing this strategy (Price Waterhouse, Quantitative and qualitative methods were 1997). (UNICEF/Ghana has been a major used to address the study’s four hypotheses. supporter of Credit with Education in Ghana Two major survey and anthropometric and of the Lower Pra Rural Bank program (heights and weights) data collection rounds in particular.) The review concluded that were carried out—a 1993 baseline and a Credit with Education, relative to the Lower 1996 follow-up—with different mother/child Pra Rural Bank’s other lines of business, had pairs participating in the two time periods. higher operational and financial self-suffi- Much of this report is dedicated to the pre- ciency ratios and efficiency (operating cost sentation of the more quantifiable findings per amount lent). The assessment concluded provided by the survey and measurements. that for the calendar year 1996, the Credit However, more qualitative techniques, such with Education program was operationally as in-depth individual or group interviews self-sufficient (interest income covered all of with participants, nonparticipants and pro- the Lower Pra’s operating costs to deliver the gram staff, were also employed and provided credit and education services), but not yet key information at each stage of the study. financially self-sufficient (covering all financial costs including imputed cost of capital Qualitative Methods—In-depth and devaluation of the loan capital due to Interviews inflation). During the baseline period, in-depth inter- Freedom from Hunger’s own assessment in- views provided rich information on women’s dicates that, at least for the last six months income-generating activities, women’s previ- of 1996, the Lower Pra program was not yet ous credit experience, and the reasons behind achieving full operational self-sufficiency. common maternal and child health/nutrition The discrepancy between the two ratios in- beliefs and practices. In the interim period dicate the difficulty of assessing a program’s between the two survey rounds, qualitative operational and financial self-sufficiency methods were used to (1) identify site-spe- within the context of a larger institution. cific manifestations of women’s empower- However, it is important to note that the ment and self-confidence, (2) more openly Lower Pra no longer receives any external explore aspects of program impact, and (3) subsidy (from Freedom from Hunger or assess the adequacy of delivery of the credit UNICEF/Ghana or any other donor) to and education services, in particular the implement the program and therefore must quality of the learning sessions designed to find the program sufficiently cost-effective motivate behavior change. Informal discus- and profitable to sustain. While not yet fully sion groups, observations of program meet- financially sustainable, Credit with Education, ings, and in-depth individual interviews with as implemented by the Lower Pra Rural field agents, participants and nonpartici- Bank, has a higher level of cost recovery than pants were undertaken. In the 1996 follow- most income-generation interventions and up period, in-depth interviews with nonpar- certainly more than traditional health/nu- ticipants as well as participants provided a trition education programs. better sense of community-level effects of the program, and interviews with borrowers without young children (the focus of the survey was on those with young children) provided a more representative view of the experience of Credit with Education participants.

Research Paper No. 4!9 Quantitative Methods—Survey and clude the same women in the two time peri- Anthropometric Measures ods, since few women had children of the desired age in both years. Baseline Data Collection The baseline survey collected information on In August 1993, a baseline survey was con- a variety of topics: ducted in 19 communities in the Lower Pra ! Rural Bank service area. (Study communi- About the household—demographics, ties were identified by the Lower Pra Rural assets, food security, food expenditures Bank as communities that were appropriate and decision making. and interested in the program but which had ! About the mother—her education, not yet been offered Credit with Education.) literacy, birth history, knowledge and Communities were classified as either being practice of key breastfeeding and “large” or “small” on the basis of whether the complementary feeding behaviors, total estimated population was greater or less diarrhea treatment and prevention, than 800 persons. Ten mother/child pairs immunizations and family planning were randomly selected to be included in the behaviors, income-earning activities, baseline survey in the “small” communities, microenterprise and wage income, and 30 mother/child pairs were randomly savings, assets and expenditures. selected in the “large” communities. Women ! About the child—breastfeeding and were randomly selected from a comprehen- eating history, estimates of diet quality sive list prepared in each community of all and quantity in the last three days and the women having a one-year-old child (12 immunization history. to 23 months of age). A total of 370 women was interviewed, although two women were In addition, heights and weights of mothers ultimately dropped when it became clear that and their children were measured to deter- their child was outside the age range. In all mine nutritional status. Portable adult/child cases, a consent form, approved by the Uni- measuring boards were used for the height versity of California, Davis, Human Subjects (length) measures. Special care was taken Research Committee, was read to all poten- to get accurate recumbent height measures tial respondents and their voluntary consent of the one-year-old child by (1) using three sought before the interviews were conducted. people, one at the child’s feet, one at the knees and one at the head, to ensure the child The one-year-old age group (12 to 23 was correctly positioned, and (2) assigning months) had been selected as the focus of only two people to take the height reading, the study, because it is among the most nu- to increase consistency of readings. tritionally sensitive and frequently malnourished age group. There is a common pattern Assignment of the Study Communities to among children in the developing world— “Program” and “Control” Samples nutritional status deteriorates beginning A quasi-experimental design was applied at when they are about 5 to 6 months through the community level to minimize possible approximately 12 months of age. Many of bias between the study groups. A common the health/nutrition education sessions con- problem in interpreting program evaluations ducted at the Credit Association meetings is the question of whether there were system- aim to prevent this predictable deterioration atic differences between the “program” and in children’s nutritional status by promot- “control” samples. It is also possible that ing good breastfeeding and complementary programs tend to be offered to the “better- feeding practices. For this reason, given the off” communities or the communities that study’s focus on assessing impact on are better organized and that more effec- children’s nutritional status and the relatively advocate for their needs. If this is the tively short duration of the study, this was case, then positive differences found between the logical age group on which to focus. How- the “participant” and “nonparticipant” ever, this decision made it unfeasible to in- groups might be due to important commu-

Impact of Credit with Education !10 nity-level differences rather than the impact sure empowerment and a few other aspects of the program. In preparation for the of program impact that emerged as impor- baseline survey, a program representative had tant from the qualitative interviews. (A copy visited each potential study community to of the English version of the follow-up sur- explain the purpose of the Credit with Educa- vey is attached as Annex A. The Fante ver- tion program and the research. Voluntary sion of the survey is also available upon re- participation in the study was sought at that quest.) time from local leaders. In each instance, it Three types of women were included in the was made clear that the community might follow-up: participants, nonparticipants in be assigned to a control group that would program communities and residents of con- not be offered the program for two years. trol communities. All participants in pro- Following baseline data collection, the 19 gram communities who had completed at study communities were stratified into four least three four-month loan cycles and had a groups on the basis of their size and access child under three years of age were included. to a main road. For each stratification, the Nonparticipants in program communities study communities were assigned to a pro- and residents of control communities were gram or control group. The majority of study randomly sampled from comprehensive lists communities, 13 of 19, were assigned ran- of all the women with children under three domly. In three cases, the Rural Bank felt years of age. In program communities, the obligated to offer the program to a particu- number of nonparticipants was selected to lar study community. For these cases, match the number of participants with chil- matched controls were selected on the basis dren of the desired age. In the control com- of their similarity to program communities munities, as with the baseline data collection in terms of proximity, commercial develop- round, either 10 or 30 residents were ran- ment, size and access to main roads. domly selected depending on whether the Communities in the program sample were of- community was classified as “small” or fered Credit with Education as soon as possible “large.” after the baseline research, while those in the Sample Size control sample were not offered the program until after completion of the research. In one A target of 360 mother/child pairs had been of the “program” communities, Dwomo, set for the baseline survey to ensure an ad- there was insufficient interest among women equate sample size to capture meaningful and to organize a Credit Association. Dwomo statistically significant differences in the nu- and only one other community, Botodwina, tritional status of children 12 to 23 months. made up the stratification “large community/ Nutritional status is measured by mean z- off main road.” For this reason, this entire scores (weight-for-age, weight-for-length and stratification was dropped from the study. length-for-age) of children from three “types” In addition, the assignments of one of these of mothers: participants, nonparticipants matched program/control pairs was ulti- living in program communities and residents mately reversed when, for political reasons, of control communities. Assuming a stan- it was not possible for the Rural Bank to be- dard deviation similar to other nutritional gin the program in one of the three commu- status studies, the necessary sample size to nities to which it had expressed an early com- detect a 0.4 difference in the z-score values mitment. of the program and control groups with a power of 0.8 and significance level of 0.05 Follow-up Data Collection (one-tailed test assuming a more favorable The August 1996 follow-up survey was con- value in the program group) would be 75. ducted in 17 of the original 19 study com- An additional five respondents per sample munities, although with different mother/ were added to compensate for possible miss- child pairs. Virtually the same survey was ing or unreliable data. The target for each used, with the addition of questions to mea- of the three groups was then increased by

Research Paper No. 4!11 50% to 120 per group to allow for more ro- “program” study communities had a total of bust sample sizes. Because two large com- 443 borrowers. Only 32, or 7%, of these munities had been withdrawn from the women had a one-year-old child and had bor- study, approximately 100 rather than 120 rowed for at least one year. While there were mother/child pairs were sampled from each more women living in households with one- of the three groups. year-olds (primarily grandmothers or aunts), For the follow-up data collection round, it for simplicity the study focused on mother/ was necessary to expand the age range to in- child pairs only. Table 2.1 also indicates that clude children under three years of age be- approximately 20%, or 90 of the total bor- cause fewer than predicted program partici- rowers, were mothers of children under three pants had one-year-old children (12 to 23 years of age, and 61% of these children were months) in 1996. Table 2.1 shows that the born after their mothers had joined the pro- 15 Credit Associations organized in the 11 gram. (The children of participants in the

Table 2.1: Credit Association Age and Membership Profile in Study Communities (11 communities w/15 Credit Associations)

# of # of Borrowers # of 3- Community Total Borrowers with a yr-olds Average # and Number with a 1- Child Born of Loan Credit Date of Current of yr-old Under 3 After Cycles Association First Loan Borrowers Child (12- Years. (<36 Mother Completed Name and I.D. Loan Cycle in the CA 23 mo.) mo.) Joined by Mother Shama Junction 5/11/93 9 31 3 8 7 6.1 Ekor Ye - 008 Shama Junction 21/7/94 7 35 2 5 3 4.3 Ntoboase - 010 Shama Junction 20/7/94 7 27 1 5 4 4.5 Boafo - 011 Atwereboanda 9/11/93 4 30 1 1 1 3 Abotar - 009 Obinyim Okyena 17/8/94 6 28 4 7 2 3.7 Odo - 027 Old Daboase 11/8/94 6 27 1 7 4 4.4 Junct. Domfo-26 Aboso 22/7/94 6 35 2 7 4 4.7 Enyidado - 018 Yabiw 26/7/94 6 34 3 7 4 5 Adom - 022 Anto 22/7/94 7 20 3 6 5 6 Bethel Nyame - 016 Anto 22/7/94 7 31 3 10 8 5.3 Ntoboase - 017 Essaman 28/7/94 7 32 1 4 1 5.5 Adom - 024 Assorkor 21/7/94 6 36 3 5 3 4.4 Anuado - 014 Assorkor 21/7/94 6 24 2 3 3 5 Ebusua - 015 Nyankrom 26/7/95 4 25 0 2 1 3 Ebenezer - 041 Beposa 7/95 4 28 3 9 5 3 Gyedzi - 036

AVERAGE TOTAL TOTAL TOTAL TOTAL AVERAGE

SUMMARY 6.1 443 32 90 55 4.5 FIGURES Cycles Borrowers (7% of All (20.3% of (61% of (18 Months) Borrowers) All Children Borrowers) Under 3)

Impact of Credit with Education !12 Table 2.2: Sample Sizes for Baseline and Follow-up Data Collection

1993 Baseline Samples 1996 Follow-up Samples

11 Study Communities Received the Program

48 “future participants” 86 participants (borrowers for at least one year)

152 “future nonparticipants” 105 nonparticipants in program communities

6 Study Communities Did Not Receive the Program 99 residents in control communities 99 residents in control communities

follow-up sample represented from 2% to fered the program. In the 11 communities 22% of all children under three years of age that did get the program, 48 women chose in the communities that received the pro- to join while 152 chose not to join—in the gram.) It was decided not to expand the age remaining 9 cases, the woman was either in- range beyond 36 months of age, since to do eligible to join, or she had died, or she could so would reduce the likelihood that the not be identified to determine her future par- mother had an opportunity to learn and ap- ticipation status. The “future participants” ply the ideal behaviors important to babies’ were all active participants at some point, good growth in their first year of life. Ap- although their duration of participation var- proximately 7% of the under-three-year-old ied, and some had left the program before children in all the study communities were the 1996 follow-up research was conducted. included in the three groups of the follow- Retrospectively classifying the baseline re- up survey. spondents by their future program participation is very helpful for dealing with the Analysis possibility of self-selection bias, which con- Program impact is evaluated by comparing founds many credit program impact evalua- the difference between the two time periods tions. “Self-selection bias” refers to the pos- (1993 and 1996) for participants, nonpar- sibility that differences found in the impact ticipants and residents in control communi- areas of concern might reflect systematic, ties. None of the 1993 respondents were pre-program differences between the women participants when the baseline was carried who join the program and those who do not, out. For the analysis, baseline respondents rather than reflecting the impact of the pro- in program communities were classified ret- gram itself. For example, if participants are rospectively by whether or not they ever found to have better nutritional status than joined the program when it was later offered nonparticipants, it is possible that this is not in their community. Consequently, baseline a result of their program participation but respondents in study communities which because women who are better nourished later received the program are classified ei- tend to join the program. By comparing the ther as “future participants” or “future non- measures of future participants (in 1993) to participants.” actual participants (in 1996), the difference

Table 2.2 summarizes the number of women between years can be attributed more reli- in each of the three groups for both survey ably to the impact of the program and not rounds. (Note that the number of mother/ to inherent differences among respondent child pairs is greater, as some women had groups. Both groups had similar inclination more than one child under three years of to join the program once it was offered. age.) Of the 368 women interviewed for the baseline, the responses of 60 women were dropped when the two “large/off main road” communities were dropped from the study. Of the remaining 308 baseline respondents, 99 lived in communities that were not of-

Research Paper No. 4!13 ate a list of consumer goods that represent a progression of wealth within the local context. All respondents were asked whether 3.0 SURVEY RESULTS: they owned ten different consumer or pro- CHARACTERISTICS OF ductive assets (radio/tape player, television, RESPONDENT SAMPLE GROUPS bicycle, water barrel, sewing machine, trac- tor, outboard motor, chain saw, motorcycle Optimally, in an evaluation of impact, the and car), and if so, how many. Respondents participant and nonparticipant (control) were also asked to estimate the current value groups would only differ in terms of their ex- of the asset by considering the price they posure to the intervention being studied. would charge if they were to sell the asset at Otherwise, if there are important differences the time of the interview. Table 3.1 summa- between the groups, it is these differences rizes the mean total value of these assets in rather than the impact of the program that 1993 dollars. Given the high degree of skew might explain contrasts in the outcome mea- and variability in these values, statistical tests sures. For this reason, it is very important to assess differences were done on logarithms to compare key socioeconomic and demo- of the measured values. Despite what ap- graphic characteristics of the sample groups pear to be relatively large differences in the which might explain the differences found in mean asset value (for example, between the the groups’ responses. control community residents and future par- By comparing the groups, the evaluation re- ticipant samples), there are no statistically search sheds light on important questions significant differences among the three that implementers have about whom the pro- groups. gram is reaching. Credit with Education was To limit the effect of the considerable vari- designed to assist poor households vulnerable ability in this measure of wealth, the dollar to hunger and malnutrition. Certain pro- value of assets was used to establish a rela- gram policies, such as selecting poor, rural tive wealth ranking. Based on the distribu- communities to implement the program and tion of the asset values, cutoff points were the relatively small loan size, make it likely determined so that households could be clas- that the program is reaching its intended cli- sified as to whether they fell in the poorest, entele. However, within a community, who poor to middle, middle to upper, or highest ultimately joins the program will depend on income quartile. For example, 25% of the the decisions taken by individual women and baseline households were classified as being the groups. With this evaluation research, in the poorest wealth quartile and assigned going back to the baseline respondents to de- a “1”—the total dollar value of their assets termine whether they had ever joined the was less than $1.40—whereas another 25% program also provided the opportunity to were classified as being in the wealthiest learn the major factors that explain why cer- quartile and assigned a “4”—the total value tain women choose not to join the program. of their assets was greater than $121.86. No statistically significant difference The mean income quartile across the three (p<0.05) was found in key household and groups was very similar and again indicated maternal demographic and socioeconomic that there was no significant difference in characteristics across the three groups: par- household income or wealth across the three ticipants, nonparticipants in program com- groups. munities and residents of control communi- In terms of mothers’ characteristics, future ties. Households were of similar size and so- participants were found to be somewhat cioeconomic status (see Table 3.1). older (p=0.1 for comparison with future Household socioeconomic status was assessed nonparticipants) and had given birth to more in two ways. First, a good proxy for income children than the two comparison groups or socioeconomic status is the value of a (p=0.06 vs. future nonparticipants and household’s assets. Program staff helped cre- p=0.07 vs. control community residents).

Impact of Credit with Education !14 Table 3.1: Baseline Survey: Household and Mother’s Characteristics Across Sample Groups—Mean (and Standard Deviation) Statistically Control Significant Program Communities Communities Difference Future Future Participants Nonparticipants Residents N=48 N=152 N=99 (p≤0.05) Household 6.4 6.3 6.7 None Family Size (2.2) (2.4) (2.9) Value of Selected $105 $176 $230 None Assets (214) (737) (741) Income Quartile 2.4 2.4 2.5 None (1.2) (1.0) (.99) Mother’s Age 29.9 27.6 27.7 None (7.5) (6.5) (6.1) Mother’s Height 156 156 157.6 None (4.3) (5.2) (5.2) Mother’s Years 5.3 4.6 4.4 None Formal Education (4.4) (4.5) (4.7) % Literate 42% 32% 32% None

Mother Earned 69% 59% 71% None Microenterprise or Wage Income in Preceding Month Mother’s Living 3.5 3.0 2.9 None Children (2.3) (1.8) (1.7) Mother’s Total 4.5 3.5 3.5 None Children (2.6) (2.1) (2) However, as the future participants had also a sketchy profile emerges of the differences lost more children, there was no significant between women who chose to join the pro- difference in the number of living children gram when it was offered and those who for the three groups. In addition, although chose not to join. Women who self-selected a higher percentage of participants was func- for the program tended to be somewhat older tionally literate (they thought they could (and perhaps as a result had given birth to read a letter if one was sent to them), this more children), to be more likely to have a difference was not significant nor was the nonfarm income-generating activity and to number of years of formal education. Fu- be somewhat more likely to be literate. ture participants were more likely to have Classifying the baseline respondents as fu- been engaged in a nonfarm income-generat- ture participants or nonparticipants also pro- ing activity at the time of the survey than vided a chance to better understand why future nonparticipants (p=0.06) but not some women chose not to join the program. more likely than residents in control communities. In early 1997, a program representative spent approximately two weeks visiting each Not surprisingly, there was no difference in of the “program” study communities to relo- the maternal heights for the three groups, a cate the baseline respondents to learn variable that was included as it is significantly whether they ever joined the program. Of related to children’s stature and subse- the 152 women who chose not to join the quently some measures of their nutritional program, 121 were directly recontacted to status. find out why. The responses are illustrative Although there are no statistically significant of the diversity of reasons women might not differences among the three baseline groups, join a Credit with Education program.

Research Paper No. 4!15 A quarter (25%) of the future nonpartici- isolation also seemed to explain why a sig- pants contacted did not join the program nificant number of women had not joined. because they were not interested in a work- Fourteen percent (14%) said that they ei- ing capital loan; either they were engaged ther had not heard about the program, had only in farming, they had no time, or they not understood the program, or were unable had no good loan investment ideas. Twenty- to organize a solidarity group. It is interest- two percent (22%) did not join the program ing that of the reasons commonly heard an- because either they had moved from the area ecdotally for why women do not join, some or they did not stay permanently in the pro- did not emerge in this survey as major fac- gram community and so were unable to ad- tors. For example, only one woman said that here to the weekly meeting requirement. she was not interested in the program be- However, 16% of the women said that “fear” cause the loans were too small, and only eight of repayment problems had kept them from women said that their husbands had either joining. Another 7% felt that weekly repay- advised or not allowed them to join. ment would be a problem for them. Social Table 3.2: Follow-up Survey: Household and Mother’s Characteristics Across Sample Groups—Mean (and Standard Deviation) Statistically Control Significant Program Communities Communities Difference Participants Nonparticipants Residents N=86 N=105 N=99 (p≤0.05) Household 6.0 5.7 6.1 None Family Size (1.9) (2.2) (2.3) Value of $125.65 $96.93 $154.18 Nonparticipants vs. Selected (270) (225) (487) participants and Assets Nonparticipants vs. controls Income Quartile 2.6 2.3 2.7 Nonparticipants vs. (1.1) (1.2) (1.0) controls Mother’s Age 33.4 27.3 28.1 Participants vs. (6.9) (6.2) (6.8) nonparticipants and Participants vs. controls Mother’s Height 157 156.8 157.6 None (5.7) (5.8) (5.3) Mother’s Years 4.4 5.0 4.0 None Formal Education (4.4) (4.7) (4.2) % Literate 33% 37% 27% None

Mother Earned 91% 50% 67% Participants vs. Microenterprise nonparticipants; or Wage Income Participants vs. in Preceding controls; and Month Nonparticipants vs. controls Mother’s Living 4 2.8 3.0 Participants vs. Children (1.7) (1.7) (1.8) nonparticipants and Participants vs. controls Mother’s Total 4.6 3.2 3.3 Participants vs. Children (2.1) (1.9) (2.1) nonparticipants and Participants vs. controls

Impact of Credit with Education !16 Certain responses indicate that Credit with since, and in part due to, a woman’s partici- Education might not be reaching some of the pation in the program. most disadvantaged women. For example, it In the following analyses of impact, it is im- seems likely that women who are relatively portant to factor in these systematic differ- poorer might be (1) most fearful of taking ences among the samples. It is interesting on debt and having repayment problems to note that the participants and the resi- that would affect the rest of the group and dents of control communities are quite simi- (2) relatively more socially isolated so that lar in characteristics that are likely to influ- they do not hear about a new program or ence children’s nutritional status or women’s are not invited to join a solidarity group un- economic capacity, such as the level of their til the groups have already formed. Still, the education and household wealth. majority of reasons given by women for not joining seem to relate more to individual work or residential patterns rather than socioeconomic status. It seems likely that the 4.0 IMPACT ON THE INTERMEDIATE wealthiest and poorest members of program BENEFITS: WOMEN’S communities are not interested or able to join the Credit with Education program. However, ECONOMIC CAPACITY the few differences between the future par- The strategy’s ultimate goals—improved ticipant and future nonparticipant groups household food security and nutritional sta- indicate that the program participants are a tus—require first that the intermediate im- representative sample of women living in pacts of poverty alleviation, empowerment these rural, coastal communities of Ghana. and behavior change be achieved at the level Table 3.2 compares the same household and of the individual borrower. For this reason, maternal characteristics for the follow-up im- qualitative and quantitative methods were pact survey samples. For interpreting the used to investigate impact on each of these findings of the impact survey, it is important three areas of intermediate benefits. This to note that participants are significantly dif- section summarizes the results pertaining to ferent from nonparticipants in program com- women’s economic capacity for poverty alle- munities and residents of control communi- viation as measured by their ties in terms of their older age, greater num- ! income; ber of children, and engagement in nonfarm ! nonfarm earnings; income-generating activity. However, in the ! contribution to total household income; follow-up period, participants may be less ! personal savings; likely to be literate and may have a lower ! entrepreneurial skill; mean number of years of formal education ! food needs expenditures; and than the nonparticipants in program com- ! household expenditures. munities (but not statistically significant). The credit and savings component of Credit Another important difference in the follow- with Education has the most direct economic up survey is that participants appear to be impact; however, nonformal education on somewhat wealthier (at least in terms of the microenterprise development as well as the value of assets) than the nonparticipants, group solidarity and support also aim to im- but not wealthier than the residents in con- prove participants’ economic returns and en- trol communities. It is possible that the rela- trepreneurial skill. tively higher socioeconomic status of participants in the follow-up survey, but not the Principal and Secondary Activities baseline, is due to the impact of the program. Some of the assets that comprise the mea- Participants and nonparticipants engage in sure of wealth, such as radio, tape player and very similar work. Tables 4.1 and 4.2 sum- water barrel, were valued at under twenty marize the principal and secondary activi- dollars and could have been purchased ties of the three 1996 survey sample groups.

Research Paper No. 4!17 Nonparticipants in program communities selling cooked food than the other two groups, are somewhat more likely to engage in farm- although there was similar representation ing as their principal occupation (56%) than across the three groups for the other enter- participants (39%) and residents in control prise areas. Very few women identified ei- communities (40%). Still, when factoring in ther salaried work or casual labor as their women’s secondary work activity, a majori- principal work activity. ty of the participants farmed. A greater per- The similarity in work patterns across the centage of participants are self-employed three samples reflects the nature of the cred- (58%) as a principal occupation compared it and loan activities characteristic of pover- with nonparticipants (34%) or control resi- ty lending programs like Credit with Educa- dents (50%). Still, 61% of the nonpartici- tion. Borrowers typically have experience in pants and 78% of the control residents had the income-generating activity for which they also engaged in a self-employed enterprise take a loan. Reliable access to credit enables over the last 12 months. In combining the women to expand their existing activities and principal and secondary activities, partici- potentially operate them more profitably and pants showed notably higher involvement in regularly. While it is not uncommon for a Table 4.1: Principal Work Activity Residents of Non- Control Participants Participants Communities Activity Type (N=84) (N=104) (N=98) FARMING 33 (39%) 58 (56%) 39 (40%)

SELF-EMPLOYED 49 (58%) 35 (34%) 59 (50%) (subtotal)

Cooked Food 20 12 8 doughnuts, tea and bread, fried polo, banku and fish, kenkey, rice and beans

Make/Sell Oil 9317

Non-Food Commerce 21 2 plastic goods, cooking utensils, used clothes

Processed Agricultural Products 25 8 gari, charcoal, cement and flour, gin, soap

Sell Fish/Lobster 92 4

Sell Foodstuffs/Agricultural Products 69 8 tabletop sales, uncooked rice, garden vegetables, maize, coconuts, oranges and bananas

Services 13 2 seamstress, hair dresser

CASUAL LABORERS 0 (0%) 2 (2%) 3 (3%) farm, nonfarm, coconut cracker, fish seller

SALARIED WORKERS 2 (2%) 4 (4%) 2 (2%) teacher, nurse, secretary, fire worker

NO WORK 0 (0%) 5 (5%) 4 (4%)

Impact of Credit with Education !18 woman to undertake a new activity or add a farm income from two different activities in new product over the course of her partici- the four weeks before the survey as compared pation, the loan activities in general reflect to only 9% of nonparticipants and 11% of the principal work traditionally undertaken the control group. Participants’ nonfarm ac- by women in the program area. tivities and earnings also seemed more sta- ble. In the four weeks preceding the inter- Participants, however, tend to be engaged in view, 91% of the participants had earned a greater diversity of income-generating ac- nonfarm income as compared to 51% of non- tivities than nonparticipants. Eighty percent participants (significantly different p<.001) (80%) of the participant sample had second- and 67% of the control group (significantly ary work as compared to only 50% of the different p<.001). nonparticipants and 60% of residents in control communities. As well, it is common for Access to Credit and Savings Services participants to engage in multiple activities at the same time. Almost a quarter (24%) As shown in Tables 4.1 and 4.2, women in of the 1996 participants had earned non- the study area are very economically active, Table 4.2: Secondary Work Activity Residents of Non- Control Participants Participants Communities Activity Type (N=83) (N=103) (N=98) FARMING 17 (20%) 15 (14%) 15 (15%)

SELF-EMPLOYED 50 (60%) 37 (35%) 41 (41%) (subtotal)

Cooked Food 16 10 8 doughnuts, fried yams/plantains, fried polo, bread, meat pie, kenkey, rice and beans

Make/Sell Oil 88 9

Non-Food Commerce 55 4 utensils and cookware, (used) clothes, slippers, firewood, shea butter and kerosene

Processed Agricultural Products 63 3 gari, gin, soap

Sell Fish/Lobster 24 5

Sell Foodstuffs/Agricultural Products 13 6 10 tabletop sales, uncooked rice, garden vegetables, maize, coconuts, oranges and bananas, sugar cane and yaka yaka

Services 01 0 seamstress

CASUAL LABORERS 0 (0%) 0 (0%) 3 (5%) farm, coconut cracker

SALARIED WORKERS 0 (0%) 0 (0%) 0 (0%)

NO SECONDARY WORK 16 (19%) 51 (49%) 39 (39%)

Research Paper No. 4!19 Table 4.3: Participants’ Program Loans and Savings Number of Amount of First Amount of Current 4-Month Loan Amount of Savings on Program Loan Program Loan Cycles Completed Deposit with Program (Mean and Range) (Mean and Range) (Mean and Range) (per Borrower) 102,062 cedis 4.7 loan cycles 12,818 50,023 cedis $60.39 approx. 19 mos $7.58 (7,000 - 150,000 cedis) (7,000 - 300,000 cedis) (3 to 8 cycles) (800 to 139,000 cedis) and despite the rural nature of the commu- Still, nonparticipants and residents in the nities, women are commonly involved in a va- control communities clearly had less access riety of nonfarm microenterprises. Women to cash credit. Only 13 (12%) of the non- frequently mentioned that what influenced participants in program communities and 16 them most as to what income-generating ac- (18%) of the residents in control communi- tivities they might undertake was whether ties had taken a cash loan to finance their they had working capital or could get the nec- current nonfarm income-generating activi- essary inputs on a credit basis. Access to cash ties. As there are virtually no other credit credit was one of the benefits of program par- programs for women in the program area, it ticipation that Credit Association members is not surprising that the majority (71%) of most commonly mentioned. loans to nonparticipants were made by family or friends at no cost. Only five nonpar- Table 4.3 summarizes the program borrow- ticipants reported taking a loan at cost: ing history of the participant sample. Ninety- three from a family member or friend, one four percent (94%) of the participants had from a coconut oil wholesaler, one from the a program loan at the time of the 1996 in- Lower Pra Rural Bank but not through the terviews (five women had borrowed for at Credit with Education program, and one from least one year but were saving only at the a husband’s employer (3,000,000 cedis at time of the interview). As required by the 10% per annum to build a building). Ex- program, all members must maintain savings cluding the rather irregular 3,000,000 cedis in their Credit Association’s account with the loan, nonparticipants in program commu- Lower Pra Rural Bank. nities had taken a mean amount of 19,270 In local currency, the mean amount of pro- cedis in cash credit, and residents in control gram loans doubled over approximately 18 communities had taken a mean of 17,600 months of program participation. However, cedis to finance their current microenterprise due to the significant devaluation of the cedi activities. (from 695 cedis to the dollar in August 1993 to 1690 cedis to the dollar in August 1996), Women’s Incomes the dollar equivalency of these loans did not In general terms, the great majority of par- increase nearly as dramatically. (Dollar ticipants (over 90%) reported that their in- equivalencies have not been provided for the comes had increased since joining the Credit first loan since its timing varied by as much with Education program, with 28% reporting as a year and a half when exchange rates ex- their incomes had increased greatly. Partici- hibited much change.) In fact, during inter- pants identified the following reasons why views, participants commonly expressed a their incomes had increased: desire for larger loan sizes and a frustration with current loan levels. Unfortunately, the ! Expanded scale of income-generating Lower Pra Rural Bank has experienced seri- activity (53%). ous liquidity constraints over the last two ! Costs reduced because no longer years which have hampered both program dependent upon getting inputs on credit expansion to new borrowers and loan size basis (36%). growth among current borrowers.

Impact of Credit with Education !20 Table 4.4: Prevalence of Entrepreneurs Getting Inputs on “Credit Basis” 1996 1996 Nonparticipants in 1996 Residents in Participants Program Communities Control Communities Inputs on Credit N=86 N=105 N=99 Engaged in one nonfarm

income-generating activity 91% 51% 60%

in last four weeks Engaged in two nonfarm

income-generating activities 24% 9% 11%

in last four weeks Of those with at least one activity, percentage that 22% 40% 51% obtained inputs on “credit basis” in last four weeks Significant difference in percentage of participants getting inputs on credit versus controls (p=.003) and participants versus nonparticipants (p=.03) but not for nonparticipants versus controls.

! Costs reduced because now able to get woman explained that when she got coco- inputs in bulk (30%). nuts on credit to make oil, she did not typi- ! Undertook new activity or new products cally pay more than the going rate. She of- (28%). ten had to go to many farmers until some- one would agree to the going rate, but at ! Sold to new customers (8%). times no one would agree. The most common effect of program partici- Given women’s limited options for cash pation was allowance for the expansion of credit, either formal or informal, the current existing activities and increased profit mar- level of women’s active and widespread in- gins. Only about a quarter (28%) of the par- volvement in commerce would not be pos- ticipants attributed their increased incomes sible without these types of arrangements. to new activities or products. During the Even after joining the program, some par- in-depth interviews, borrowers commonly ticipants continued to get a portion of their spoke of the improved profit margin they inputs on credit because their program loan were able to earn after borrowing from the was not large enough to finance the scale or program. Before joining, women often got variety of activities in which they were en- all or part of the inputs they needed on credit gaged. Still, despite their greater tendency because they lacked working capital. Oper- to engage in more than one income-generat- ating a microenterprise on what is called ing activity, the 1996 program participants “credit basis” is very widespread in the pro- were significantly less likely to obtain their gram area. For example, a farmer will pro- inputs on credit than nonparticipants and vide coconuts or palm nuts to oil producers residents in control communities (see Table who pay for these inputs after selling to the 4.4). oil wholesalers coming from Accra; or a fisher may give a woman fish on credit that she Nonfarm Income Earnings will pay for after smoking the fish and sell- Women’s income was quantified by focus- ing it in a nearby market. ing on nonfarm earnings in the four weeks Information from the baseline survey indi- preceding the survey. While nonfarm in- cated that the effective interest rate in this come is likely to represent only a portion of type of arrangement was on average 17% for returns to women’s overall productive labor, a two-week period, or approximately 442% it was decided that total income estimates per annum. In some cases, there is no addi- would be too difficult and time-consuming tional cost for receiving inputs on credit; to collect. Since the program is most likely however, such arrangements depend on fa- to affect nonfarm income earnings, efforts miliarity or social ties and may involve were made to quantify this type of income. greater “search costs.” For example, one In addition, there is some evidence that it is

Research Paper No. 4!21 this type of steady earnings that most directly for some women the concept of “profit” was affects basic need expenditures such as food. more similar to the business concept of “sav- Still, even with this more narrow focus it was ings.” Before estimating their “profit,” some acknowledged that for a variety of reasons women would subtract from their earnings there were likely to be considerable error and not only their business-related costs but also variability in this measure. Because few the amount they had spent on food and other women kept accounting records, the recall household expenses. In the follow-up sur- period was limited to the past four weeks vey, a supplementary question was added to before the survey. Field agents also rated probe whether women had estimated their women’s ability to recall this information. “profit” after deducting amounts spent on food and other family expenses. The results Estimated “Profit” in 1996 are similar even after adding any amount deducted for food, etc. (see Table Women were first asked to simply estimate 4.6). their profit over the past four weeks for their nonfarm income-generating activity earnings It can be calculated from Table 4.6 that 45% for the time period that best captured the of the women in all three sample groups to- product cycle—per day, per week, per two gether had provided “profit” estimates after weeks or per month. Mean monthly profit deducting what they had spent on food. This estimates are summarized in Table 4.5. (If a tendency has a number of implications. First, woman had more than one activity, her profit it demonstrates that even if there is a com- estimates were summed.) There was a sig- monly accepted word for “profit,” people’s nificant difference in the logarithm values of concept of this rather specific business term monthly estimated profit between years for may differ resulting in systematic under- or participants versus nonparticipants and the over-reporting. (Certainly, women’s concep- control sample but not for nonparticipants tion of “profit” does not correspond with the versus controls. As compared to baseline specific business definition which requires net- measures, the increase in estimated nonfarm ting out nonmonetized business-related costs monthly profit was approximately $29 for such as depreciation of assets or wages to fam- participants, $11 for nonparticipants and ily workers.) Second, this conceptualization of $13 for the control group. (Mean amounts “profit” highlights the considerable integration are presented in 1996 dollars, controlling for and interaction of the welfare of women’s busi- United States inflation between the two time nesses to the welfare of their families. Women periods.) engage in their income-generating activities so that they can better feed and care for their Asking women to estimate their profit posed families. Credit with Education was designed to little problem for most women. Although strengthen women’s economic capacity to in- they did not keep written records, the con- vest in their families, and the tendency to fuse cept of their “profit” or their “benefit” was business and family expenses is to some degree very clear to women, and there is a well- evidence of the appropriateness of this design. known and accepted Fante word for “profit.” However, during in-depth interviews over the Table 4.7 shows an alternative measure of course of the research, it became obvious that nonfarm income earnings or estimated net

Table 4.5: Estimated “Profit” from Microenterprise or Wage Income in the Preceding Four Weeks—Mean (and Standard Deviation) Year Program Communities Control Communities

1993 Future Participants N=48 Future nonparticipants N=151 Residents N=97

$5.90 (7.3) $4.17 (7.3) $5.25 (7.3) 1996 Participants N=86 Nonparticipants N=104 Residents N=99 $34.53 (34.3) $15.47 (29.6) $18.35 (41.4) Significant difference in log value of respondent-estimated monthly profit between years for participants versus controls (p=.002) and participants versus nonparticipants (p=.001) but not for nonparticipants versus controls.

Impact of Credit with Education !22 profit. Women were asked to itemize their ations, with some women enjoying consider- business costs (not including family labor, de- able improvement in their activities while preciation or interest payments) and esti- others experienced little change. A better un- mate revenue earned over the past four derstanding of the factors that allow some weeks. Again Table 4.7 indicates there was women to be relatively more successful—in- a significant difference between years for par- dividual attributes, entrepreneurial skill, spe- ticipants versus nonparticipants and partici- cific income-earning activities or program pants versus the control sample but not for loan terms—could lead to changes in pro- nonparticipants versus controls. As com- gram implementation, such as microenter- pared to baseline measures, the increase in prise development education, which might net nonfarm monthly income was $36 for enhance the economic impacts for other, less participants, $18 for nonparticipants and successful borrowers. Box 4.8 includes ex- $17 for the control group. The net income cerpts from interviews with participants who estimates were higher than the estimated enjoyed particular success and highlights in- profit, which might further reflect the ten- vestment strategies and borrower attributes. dency of some women to net out family ex- Box 4.9 summarizes an example of the expe- penses when estimating their profit. rience of one of the Credit Associations in the study and the vulnerabilities and diffi- While overall the 1996 participants exhibit- culties its members faced in the previous loan ed significant improvement in their nonfarm cycle before the follow-up survey. earnings, it is important to note that there was considerable range in participants’ Contribution to Total Household monthly earnings. Some participants had Income net monthly nonfarm income as high as $200 to $330 per month, while 10% of the partic- An indirect effect of women’s increased cash ipants reported net earnings of $10 or less earnings is believed to be an enhancement per month. As with other impact evalua- in women’s bargaining power within the tions, it is clear there is a great diversity of household. If women have different spend- impact even within the same Credit Associ- ing preferences than men and place greater

Table 4.6: Estimated “Profit” Plus Any Amount Deducted for Food and Other Family Expenses—Mean (and Standard Deviation) Year Program Communities Control Communities 1996 Participants Nonparticipants Residents

N=85 N=101 N=97

$38.88 (36.0) $19.52 (40.8) $23.18 (53.2) % Estimating “profit” after deducting for 64% 60% 52% food Significant difference in log value of respondent-estimated monthly profit between years for participants versus controls (p<.001) and participants versus nonparticipants (p<.001) but not for nonparticipants versus controls.

Table 4.7: Net Income from Microenterprise or Wage Income in the Preceding Four Weeks—Mean (and Standard Deviation) Year Program Communities Control Communities 1993 Future Participants Future Nonparticipants Residents

N=44 N=145 N=94

$5.75 (8.2) $4.05 (8.9) $5.98 (11.1) 1996 Participants Nonparticipants Residents N=86 N=103 N=99 $41.50 (46.8) $21.95 (43.5) $22.85 (52.9) Significant difference in log value of monthly net income between years for participants versus controls (p=.001) and participants versus nonparticipants (p<.001) but not for nonparticipants versus controls.

Research Paper No. 4!23 value on nutritionally beneficial expenditures households’ total income. In the baseline such as food or health care, as is believed, an period, 26% of the “future participants” said increase in their intrahousehold bargaining either they earned no income or that it was power could have a greater impact on total only a small portion of total household in- household spending than the specific amount come. However, in 1996 only 10% of the of her increased earnings. participants gave these responses as com- Respondents in both the baseline and fol- pared to 32% of the nonparticipants in pro- low-up data collection rounds were asked to gram communities and 27% of the residents rank their relative contribution to their in control communities.

Box 4.8: Successful Borrowers

Lessons from Her Grandmother Ama Tawiah (no real names are used in these examples) owns a successful chop bar (roadside restaurant). She is 29 years old, married and has four children. Before joining the program, she like many other women sold kenkey and fish from a table. However, she wasn’t making enough profit, so with her initial 50,000 cedis program loan and another 30,000 cedis that her husband had provided, she rented a shed and switched to making fufu (a starchy staple served with soup). When she was a child, Ama lived with her grandmother, who had a fufu business. From her grandmother, she learned that to be successful you need (1) to do hard bargaining, (2) to know the quantity and what to charge in order to make a profit, (3) to keep track of whether you are making a profit or not, and (4) to put some profit aside so when an opportunity presents itself to get inputs at a good price, you are able to take advantage. Ama and other women have also been discussing good business skills during their Credit Association meetings. With this knowledge, as well as the loan capital, she has used the profits from her chop bar to expand her business. She started on a small scale and had only one laborer but now has five. She is able to buy the inputs she needs in bulk with cash and then keep them in a freezer that she was able to buy. With the profits from her business, she has even been able to help her husband establish a corn mill. Her last loan was for 250,000 cedis, and she has 54,000 cedis in her current savings. Her main suggestion for the program is that she be allowed to take a larger loan.

A Unique Product Over five loan cycles, Christina Aidoo has been able to expand her coconut biscuit business and now has considerable profits. Christina started this enterprise about six years ago when she needed a way to earn income at a time when she didn’t have any money. The biscuits were her own unique creation; there is nothing else like them on the market. She taught herself how to make them by experimenting with ingredients and then conducting small-scale market research in order to get customer reaction. Christina is 44 years old and married. She has given birth to six children, five of whom are still living. In the beginning she had one woman selling for her, but now she has four. Each woman collects 50,000 cedis worth of biscuits and is paid 10,000 cedis worth of biscuits per week. Now she doesn’t have to get inputs like flour and sugar on credit. (Flour costs 52,000 cedis per bag with cash and 55,000 cedis on credit.) Before joining the program she used one bag of flour per week, but now she needs and can afford four bags. Not having to take supplies on credit represents a significant improvement to her business profits. Christina estimated her net profit for the previous month at 240,000 cedis. During the most recent loan cycle, she borrowed 250,000 cedis which she has reinvested into her business. She also has managed to build a savings of 86,100 cedis. In the future Christina would like to build a house with the income her business and savings have provided. When asked if there was anything she would change about the program, her only response was a request for a larger loan size.

Impact of Credit with Education !24 Entrepreneurial Skill ever, it is also recognized that Credit with Edu- How the program loan is invested will greatly cation is serving women operating in a sur- influence the return and economic benefits vival economy, many of whom might benefit a borrower is able to enjoy. A basic assump- from practical entrepreneurial and credit-use tion of the Credit with Education strategy, like skills development. One characteristic asso- other minimalist credit approaches, is that ciated with what has been referred to as “C- the borrowers know best (or at least better level enterprises,” or pre-entrepreneurial than an external lender) what activity would microenterprises, is that the producer focuses be most profitable for them given their per- more on “supply” than “demand” consider- sonal considerations and trade-offs. How- ations. For example, the entrepreneur might

Box 4.9: Struggling Borrowers The “Hungry Season,” Sickness and Death

The Yabiw Credit Association faced difficulty repaying their fifth program loan. Because of their repayment problems, many of the members of the Yabiw Credit Association chose not to take a loan during the Association’s sixth cycle but to continue participating in the program’s education sessions and as savers. Yabiw is a relatively remote community whose inhabitants are primarily dependent on agricultural production for their well-being. The fifth loan cycle ended in June which corresponded with the “hungry season.” It is during the rainy months of April through June that people typically need to buy more food since their own cassava is inedible. Cassava, the local staple, is stored in the ground and becomes waterlogged or spoiled during the early rains. Borrowers have more difficulty making their loan repayments and savings deposits when they are more dependent on purchased food. In addition, several of the members of the Yabiw Credit Association experienced particular repayment problems due to illness or death. One such member is Leticia Mensah who has been a member of the Credit Association since its inception and who lost her husband during the fifth loan cycle. Normally, she makes and sells gari (processed cassava), but as is the custom of the area she will not work for a period following her husband’s death, and so she is not currently taking a loan. Repayment was difficult during the last cycle, but her in-laws helped her. She has four children, two of whom are under five years of age. She is 38 years old and plans to take a loan during the next loan cycle. Dorothy Cobinah became sick during the last loan cycle with serious toothaches and a swollen mouth that has not yet fully healed. Her loan activity was to sell cooked food (rice and corn water). When she became ill, she was not able to repay her loan, and so the group paid for her. She still owes the group 20,000 cedis which she will repay without interest. Because she is not currently eligible for a loan, she has only been able to engage in farming. She has been making savings deposits with the group, but it is difficult. If she is not able to harvest food from her farm, she goes to sleep without eating because she doesn’t have money to buy food. Now she is unable to get anyone to give her a loan but knows that when she does get a loan she will use it to earn profit. She is 45 years old, has three children and is divorced. Although these women have experienced serious hardships, they are still members of the program and have received assistance as a result of their participation. However, it seems that illness and death are major factors explaining why some women experience repayment problems and ultimately leave the program. In one study community, an older women was wearing virtual rags, and her young children had the telltale orange and brittle hair of malnourished children. The woman was asked if she was a member of the Credit with Education program, and if not, why not. The woman explained that she had been a member, but during her second loan cycle, one of her children fell ill. She was not able to work in order to care for her child. She also used much of her loan money to seek treatment and buy medicine for the child. Ultimately, the child died, and she then had to bear the cost of the funeral. She has no husband to help her, and she has other children to care for as well. Initially, it was a neighbor who had invited her to join the program and who encouraged the group to admit her. The Credit Association held this neighbor responsible for the older woman’s debt; they said it was her idea that the woman join. The neighbor explained that she knew this woman was very poor, and this is why she had thought the program would be good for her. However, even before her child fell ill, the woman had difficulty making her weekly repayment. Because the group held the neighbor responsible for the woman’s debt, the neighbor had to appeal to her husband, who harvested a sugarcane plot to repay the loan.

Research Paper No. 4!25 Table 4.10: Factors Women Considered When Selecting an Income-Generating Activity Participants Nonparticipants Control 1993 1996 1993 1996 1993 1996 N=48 N=86 N=152 N=105 N=98 N=97 Familiarity with the work (have done this work before, it 58% 49% 62% 43% 46% 42% is the season/others doing it) Working capital (whether have enough, whether 46% 49% 40% 51% 43% 47% work requires little capital, etc.) Time required/compatibility with other work or family 10% 7% 5% 2% 7% 8% responsibilities Demand for product/profitability of 33% 72% 38% 53% 43% 60% activity

Significant difference in percentage mentioning “demand” or “profitability” considerations between years for participants versus controls (p=.03) and participants versus nonparticipants (p=.02) but not for nonparticipants versus controls. be primarily influenced by what enterprise significant investments in quality of life im- to pursue on the basis of his or her own fa- provements. While it is true that the pro- miliarity or seasonality of this work rather gram requires some amount of mandatory than whether it is in demand or likely to yield savings, this amount can be quite nominal, profitable returns. Through loan feasibility and members are able to withdraw their sav- assessments and nonformal learning sessions ings at the end of a loan cycle. Given this about profit, sales and management strate- mandatory requirement, it is not surprising gies, the Credit with Education program aims that there was a significant difference in the to develop this type of entrepreneurial per- percentage of participants having savings be- spective. tween years versus controls and versus non- Table 4.10 summarizes the most common participants but not for nonparticipants ver- factors respondents mentioned when asked sus controls (see Table 4.11). However, there what they consider when deciding what in- was also a significant difference in log value come-generating activity to engage in. In the of amount saved between years for partici- baseline period, the most common reason pants versus controls and participants ver- given by women in each group related to sus nonparticipants but not for nonpartici- their familiarity with the activity. The sec- pants versus controls. The amount partici- ond most common factor was also a “supply” consideration—whether they had ad- pants had in savings also varied tremen- equate working capital—which indicates the dously—from $0 to $118—again indicating importance of the strategy’s credit compo- the range in economic success among those nent. While profitability or demand con- borrowers living in the same communities and siderations became more prevalent in the participating in the same program. 1996 follow-up survey for all the groups, the difference between years was only significant Expenditures for participants versus controls and partici- A major assumption underlying the design pants versus nonparticipants but not for of the Credit with Education strategy is that if nonparticipants versus controls. women are assisted in earning increased in- Savings comes, they will invest their increased profits in nutritionally beneficial items such as Savings is an important economic resource food, health, shelter and other basic needs for coping with family emergencies, develop- like clothing. In addition, it is hoped that ing an income-generating activity or making the strategy’s education component will in-

Impact of Credit with Education !26 Table 4.11: Personal Cash Savings Participants Nonparticipants Control 1993 1996 1993 1996 1993 1996 N=46 N=86 N=149 N=102 N=92 N=98 Percentage who report having 51% 99% 41% 42% 35% 49% personal savings Mean amount of $15.49 $18.24 $9.57 $12.48 $9.54 $8.73 savings (24.4) (26.1) (20.2) (28.2) (20.7) (16.9) Significant difference in percent having savings between years for participants versus controls (p<.001) and participants versus nonparticipants (p<.001) but not for nonparticipants versus controls. Significant difference in log value of amount saved between years for participants versus controls (p=.04) and participants versus nonparticipants (p=.008) but not for nonparticipants versus controls. crease awareness and appreciation for nu- household expenditure on many types of tritionally beneficial expenditures and make food. These estimates have been converted these investment decisions more likely. into dollars and, to obtain per capita amounts, divided by the number of persons Food Expenditures in the family (counting an adult as one and Table 4.12 summarizes household food ex- a dependent under 17 years of age as 0.75). penditure information. In both time peri- In the five categories of food, there was not ods, respondents were asked to estimate the much difference between years in the amount Table 4.12: Per Capita Food Expenditures in U.S. Dollars—Mean (and Standard Deviation) Participants Nonparticipants Control 1993 1996 1993 1996 1993 1996 N=48 N=90 N=152 N=112 N=96 N=102 Maize $0.26 $0.47 $0.26 $0.52 $0.32 $0.43 (.25) (.41) (.26) (.48) (.36) (.46)

Significant difference in log per capita maize expenditures between years for nonparticipants versus controls (p=.05) but not for participants versus nonparticipants or for participants versus controls. Tubers $0.25 $0.30 $0.21 $0.24 $0.22 $0.24 (.32) (.44) (.30) (.31) (.28) (.27)

Vegetables $0.20 $0.21 $0.19 $0.26 $0.23 $0.24 (.12) (.16) (.17) (.21) (.27) (.21)

Significant difference in log per capita vegetable expenditures between years for participants versus nonparticipants (p=.05) but not for participants versus controls or for nonparticipants versus controls. Meat/Fish $0.52 $0.56 $0.50 $0.61 $0.49 $0.59 (.35) (.51) (.38) (.39) (.36) (.45)

Cooked Food $0.32 $0.21 $0.35 $0.23 $0.32 $0.21 (.34) (.32) (.36) (.23) (.57) (.28)

Total $1.54 $1.75 $1.53 $1.96 $1.58 $1.78 (.85) (1.1) (.88) (1.0) (1.0) (1.1)

Marginally significant difference in log total per capita food expenditures between years for participants versus nonparticipants (p=.09) but not for participants versus controls or for nonparticipants versus controls. % Household food 50% 60% 45% 45% 51% 47% expenditures paid for by women’s income Marginally significant difference in women’s relative contribution to total per capita food expenditures, controlling for marital status, between years for participants versus controls (p=.09) but not for participants versus nonparticipants or for nonparticipants versus controls.

Research Paper No. 4!27 Table 4.13: Spending on School Fees and Materials in Last 12 Months Participants Nonparticipants Control N=86 N=105 N=99 % Women spending their own money 54% 31% 42% Controlling for number of living children, no significant difference among the three groups. Amount Spent $6.40 $5.75 $7.36 Mean (and Standard Deviation) (12.1) (15.9) (25.6) Controlling for number of living children, no significant difference among the three groups. spent across the three groups. In fact, the scope of this evaluation to determine what nonparticipants in program communities indirect effect the program might have had showed the most dramatic increases in per on total household expenditures, if indeed capita food expenditures between the two participants’ earnings “freed up” income time periods as compared to the other two earned by other members of the participant groups. In statistical tests on the logarithm households. values, there was a significant and positive difference in the per capita amount spent on Household Expenditures maize by nonparticipants versus residents in Women were asked to report how much of control communities. Nonparticipants also their own money they had spent on several had a significant and positive difference types of common household expenditure. (comparing log values) of per capita amount Their responses were analyzed both in terms spent on vegetables versus participants and of mean amounts spent on each type and a marginally significant difference for total also the prevalence of women who spent per capita food expenditures versus partici- money they had earned on that type. pants. A greater percentage of participants reported The fact that a great majority of households spending their own money on school fees and included in the survey are farming house- expenses in the last year than the other two holds that provide a significant portion of nonparticipant groups. (See Table 4.13.) their own food consumption complicates in- However, as shown in Table 3.2, participants terpretation of food expenditure informa- had a greater number of living children. tion. It is possible that the higher mean val- When controlling for the number of children, ues for nonparticipants indicates greater de- there was no statistically significant differ- pendency on purchased food rather than ence among the groups in the prevalence of higher per capita food consumption. women spending their own money on edu- There was some indication of substitution cation-related expenses. Again controlling of responsibility for food purchases within for number of children, the log value of the the participant households, which might un- amount women spent on education was not dermine the program’s impact on household significantly different. food expenditures. In the baseline period, For each of the three groups, women were 45% to 51% of total household food expen- less likely to have spent their own money on ditures were paid from income women had housing improvements in the last 12 months earned and controlled. In the follow-up pe- than on education. (See Table 4.14.) There riod, there was an increase to 60% among was no statistically significant difference in participant households, but no increase in either the prevalence or log value of spend- the other two groups. When controlling for ing on housing improvements among the marital status, this increase is not statisti- three groups. cally significant. It is possible that as participants earned more profit from their loan- Close to three-quarters of the women in each assisted activities, husbands and other mem- of the three groups reported spending some bers of the household contributed less money amount of their own money on clothing for for food purchases. It was not within the themselves in the last 12 months. (See Table

Impact of Credit with Education !28 Table 4.14: Spending on Housing Improvement in Last 12 Months Participants Nonparticipants Control N=86 N=105 N=99 % Women spending their own money 24% 16% 21% No significant difference among the three groups. Amount Spent $5.05 $2.00 $3.54 Mean (and Standard Deviation) (13.2) (5.8) (14.8) No significant difference among the three groups. 4.15.) There was no significant difference participants. (See Table 4.17.) However, in either the prevalence or the log value of there was no significant difference between spending among the three groups. the log values of the amount spent among the groups. Residents in control communities were significantly more likely than nonparticipants There was no significant difference among to have spent their own money on clothing the three groups in either prevalence of for their children even when controlling for spending their own money on medical costs the number of children. (See Table 4.16.) in the last three months or, when controlling Similarly, there was a marginally significant for the size of the family, the log values of difference for participants versus nonpartici- spending on medical costs. (See Table 4.18.) pants. However, when controlling for the Despite participants’ more active involve- number of living children, there was no sig- ment in nonfarm income-generating activi- nificant difference in the log value of spend- ties, they were no less likely than the two non- ing on children’s clothes among the three participant groups to have spent their own groups. money on agricultural inputs and/or labor Participants and residents in control com- in the six months preceding the survey. (See munities were significantly more likely to Table 4.19.) Moreover, there are no signifi- have spent their own money on business as- cant differences in the log value of amount sets in the last 12 months than were non- spent on agriculture among the three groups. Table 4.15: Spending on Clothing for Self in Last 12 Months Participants Nonparticipants Control N=86 N=105 N=99 % Women spending their own money 74% 72% 83% No significant difference in percent spending among groups. Amount spent $26.13 $29.21 $30.36 Mean (and Standard Deviation) (30.4) (35.5) (28.5) No significant difference in log value among the groups.

Table 4.16: Spending on Clothing for Children in Last 12 Months Participants Nonparticipants Control N=86 N=105 N=99 % Women spending their own money 85% 71% 84% Controlling for number of living children, marginally significant difference between participants and nonparticipants (p=.06) and significant difference for controls versus nonparticipants (p=.04) but not between participants and controls. Amount Spent $18.08 $15.87 $12.92 Mean (and Standard Deviation) (22.8) (27.7) (12.3) Controlling for number of living children, no significant difference of log value among the groups.

Research Paper No. 4!29 Table 4.17: Spending on Business Assets in Last 12 Months Participants Nonparticipants Control N=86 N=105 N=99 % Women spending their own money 45% 30% 43% Significant difference in percent spending between participants and nonparticipants (p=.04) and marginally significant difference for controls versus nonparticipants (p=.06) but not for participants versus controls. Amount spent $6.52 $4.81 $5.12 Mean (and Standard Deviation) (10.5) (16.3) (10.1) No significant difference of log value among the groups. Table 4.18: Spending on Medical Costs in Last Six Months Participants Nonparticipants Control N=86 N=105 N=99 % Women spending their own money 79% 70% 79% No significant differences among the groups. Amount spent $11.70 $10.01 $9.40 Mean (and Standard Deviation) (20.4) (22.4) (12.6)

Controlling for family size, no significant difference of log value among the groups. An income contribution scale was derived Credit with Education program. Most com- with “1” representing the greatest relative monly, participants attributed this improve- contribution (100% or “all of the household ment to expansion of their businesses, re- income”) and with “6” representing the low- duced input costs by buying in bulk or with est relative contribution (0% or “earned no cash, and new activities or products that ac- income”). The mean value for the 1996 par- cess to credit had made possible. There was ticipants is a significantly higher propor- a significant and positive difference in in- tional contribution than the other two crease of participants’ monthly nonfarm groups. (See Table 4.20.) However, in terms earnings as compared to nonparticipants and of difference between the baseline and fol- residents in control communities. As com- low-up periods, there was no significant dif- pared to baseline measures, the increase in ference among the three groups. net nonfarm monthly income (revenue mi- Conclusion nus costs) was $36 for participants, $18 for nonparticipants and $17 for residents in con- Participants, nonparticipants in program trol communities. While overall the 1996 communities and residents in control com- participants exhibited significant improve- munities engage in very similar work. The ment in their nonfarm earnings, there was majority of women in each group farmed and considerable range in monthly profit even in had undertaken at least one nonfarm enterprise in the last year. The most common the same Credit Associations. Some partici- microenterprises were selling cooked food, pants had net monthly enterprise incomes making and selling coconut or palm oil, sell- as high as $200 to $300 per month, but 10% ing fish or trading foodstuffs such as veg- had net incomes of $10 or less. etables, maize or rice. Participants engaged Although nonfarm incomes increased, the in a greater diversity of enterprises with increase in participants’ assessment of their seemingly greater regularity of operations relative contribution to their households’ to- relative to the two nonparticipant groups. tal incomes was no greater than in the two The majority of the 1996 participants (90%) nonparticipant groups. There were also few felt that their incomes had “increased” or “in- differences across the groups in household creased greatly” since they had joined the expenditures.

Impact of Credit with Education !30 Table 4.19: Spending on Agricultural Inputs and Hired Labor in Last Six Months Participants Nonparticipants Control N=86 N=105 N=99 % Women spending their own money 48% 43% 47% No significant differences among the groups. Amount spent $4.30 $3.61 $2.85 Mean (and Standard Deviation) (7.0) (7.1) (5.8) No significant differences in log values among the groups.

Table 4.20: In the Last Year, Relative Contribution of Respondents’ Income to Total Household Income Participants Nonparticipants Control 1993 1996 1993 1996 1993 1996 N=46 N=86 N=149 N=102 N=92 N=98 All of household income 17% 12% 4.0% 3% 11% 5% Most of household income 26% 20% 15% 13% 8% 15% Half of household income 9% 29% 19% 22% 24% 22% Some but less than half of household income 22% 29% 25% 30% 23% 30% Very small portion of household income 24% 10% 32% 22% 31% 18% Earned no income 2% --- 4% 10% 3% 9% Contribution scale (1=all; 6=none) Mean (and Standard 3.2 3.0 3.8 3.9 3.7 3.7 Deviation) (1.5) (1.2) (1.3) (1.3) (1.4) (1.3)

However, there was evidence that the pro- Still, approximately a quarter of the 1996 gram was fostering the entrepreneurial skills participants reported facing some difficulty of participants, who were significantly more making their weekly or end-of-cycle loan re- likely to consider demand and profitability payments. Illness or a death in the family when deciding to invest in income-generat- was the root cause of many of the few seri- ing activities than were nonparticipants and ous repayment problems borrowers faced. residents in control communities. There was also a significant difference in savings held by participants versus controls and participants versus nonparticipants.

Research Paper No. 4!31 Colostrum For the three baseline groups, between 35% 5.0 IMPACT ON THE INTERMEDIATE and 40% of the mothers discarded their co- BENEFITS: WOMEN’S HEALTH/ lostrum either before or after their child was born. Informal discussion groups did not re- NUTRITION KNOWLEDGE AND veal any strong cultural taboo against giving PRACTICE infants colostrum. Rather, some mothers An assumption underlying the design of the simply believed its thick, yellowish appear- Credit with Education strategy is that while ance meant colostrum was unhygienic and poverty is the root cause of malnutrition, in- not good for the baby. Through skits and come increases alone are unlikely to posi- group discussions, field agents facilitated tively influence children’s nutritional status. learning sessions that emphasized the ben- Key health and nutrition practices that will efits of colostrum for keeping newborns best nourish children, keep them healthy healthy. and promote their good growth are also im- To evaluate whether program education had portant. Hence, nonformal health and nu- impacted this behavior, it was necessary to trition education is directly integrated into restrict the participant group to those women the credit delivery system and the learning who would have had an opportunity to try sessions facilitated at the Credit Associations’ what they had learned—women who had regular meetings. given birth after joining the Credit with Edu- The evaluation collected information on cation program. Figure 5.1 shows that while knowledge and practice in each of the only 60% of the 1993 future participants program’s health/nutrition topic areas: had given their newborns colostrum, 98% of the 1996 participants who gave birth after ! Breastfeeding. joining the program had done so. There were ! Child feeding. significant and positive differences between ! Diarrhea treatment and prevention. years for participants versus controls and for ! Immunization. participants versus nonparticipants, indicat- Program impact is assessed through coming a positive effect of the program on en- parisons of baseline and follow-up measures couraging mothers to give newborns colos- of women’s knowledge and practice of the trum. specific ideal behaviors promoted in each of these topic areas. In addition, to better ex- Exclusive Breastfeeding plore the relationship between the desired An ideal behavior that received considerable intermediate benefits and children’s nutri- attention by the program was exclusive tional status, a composite score is derived for breastfeeding (meaning no water or foods) each topic area. until babies are approximately six months of Breastfeeding Promotion age. The baseline research found that a great majority of mothers (93%) introduced wa- The ideal breastfeeding behaviors promoted ter during their newborn’s first week of life. by the program include: (1) giving the child Virtually all mothers believed newborns the first antibody-rich milk, colostrum, needed water to survive. However, even in rather than discarding it; (2) exclusively hot, dry climates, breastmilk contains suf- breastfeeding babies until they are approxi- ficient water for a young baby’s needs. Giv- mately six months of age; (3) not using feeding them water greatly increases their risk ing bottles; and (4) breastfeeding until the of getting diarrhea and other illnesses. In child is approximately two years of age. terms of “first foods,” 90% of the baseline respondents had introduced watery foods (like koko, a maize-based porridge, or mashed kenkey, a maize-based polenta) by

Impact of Credit with Education !32 the end of the baby’s fourth month (120 Associations made a positive example of days). these early innovators by awarding them cer- tificates and keeping snapshots of partici- Through role plays and group discussions, pants’ “plumpy” babies that had been ex- field agents emphasized that breastmilk clusively breastfed for about six months. alone is the safest and best food for babies and provides all the necessary nutrition and Again, the participant sample was restricted hydration a baby requires from birth through to include only those women who gave birth about six months. During the period of the after joining the program and who had the research, this same message was increasingly opportunity to apply program messages. Fig- being promoted by local health centers due ures 5.2 and 5.3 demonstrate the remark- to training provided by UNICEF to health- able change in exclusive breastfeeding prac- care providers. Reinforcement from multi- tices in general, but most dramatically for ple, trusted sources was probably instrumen- Credit with Education participants. There tal in convincing participants to try this rev- were significant differences in the mean age olutionary behavior. Several of the Credit when water and watery foods were first in-

Figure 5.1: Gave Colostrum to Newborns

120% 98% 100% 78% 80% 71% 60% 65% 61% 60% 40% 20% 0% % Respondents Progr am Communities Progr am Communities Control Communities Future Future 1993 1996 Participants Participants Nonpartic. Nonpartic. (n=92) (n=102) 1993 1996 1993 1996 (n=47) (n=55) (n=144) (n=112)

Comparing 1993 and 1996 *Significant difference for participants versus control group (p<.001). *Significant difference for participants versus nonparticipants (p=.006). *No significant difference for nonparticipants versus control group.

Figure 5.2: Mean Age Water Introduced to Babies

150 125 120

90 63 51 60 30 7 4 8 Age in Days 0 Program Communities Program Communities Control Communities Future Future 1993 1996 Participants Participants Nonpartic. Nonpartic. (n=93) (n=97) 1993 1996 1993 1996 (n=45) (n=46) (n=147) (n=106)

Comparing 1993 and 1996

*Significant difference for participants versus control group (p<.001). *Significant difference for participants versus nonparticipants (p<.001). *No significant difference for nonparticipants versus control group.

Research Paper No. 4!33 Figure 5.3: Mean Age Watery Foods Introduced to Babies

6 5 5 4.1 4 3.5 2.9 3 2.2 2.3

Months 2 1 0 Program Communities Program Communities Control Communities Future Future 1993 1996 Participants Participants Nonpartic. Nonpartic. (n=87) (n=90) 1993 1996 1993 1996 (n=45) (n=44) (n=134) (n=103)

Comparing 1993 and 1996

*Significant difference for participants versus control group (p=.004). *Significant difference for participants versus nonparticipants (p=.009). *No significant difference for nonparticipants versus control group. troduced to newborns between years for both ever using a bottle to feed their under-three- participants versus residents of control com- year-old babies. However, the decline in munities and for participants versus nonpar- bottle use (from 88% to 23%) was most dra- ticipants. These graphs indicate the positive matic for 1996 participants. (Since bottle and significant impact the program had on use is likely to increase with the age of the exclusive breastfeeding from women who gave child, for this analysis, all participant chil- birth after joining the Credit with Education dren were included and not only those born program. On average, participants did not after their mother had joined the Credit with introduce water or watery foods until their Education program.) A positive and signifi- babies were in their fifth month of life. cant impact of the program was evident in reduced frequency in feeding bottle use by Feeding Bottle Use participants versus controls and participants Figure 5.4 also indicates area-wide improve- versus nonparticipants when controlling for ments. Fewer women in each group reported the child’s age. No significant difference was Figure 5.4: Ever Used Bottle to Feed Baby

120% 100% 88% 88% 83% 80% 60% 60% 52% 40% 20% 23% 0% Pr ogram Communities Program Communities Control Communities Future Future 1993 1996 % One-Year-Old Children Participants Participants Nonpartic. Nonpartic. (n=99) (n=103) 1993 1996 1993 1996 (n=48) (n=90) (n=151) (n=109)

Comparing 1993 and 1996

*Significant difference for participants versus controls, controlling f or child’s age (p<.001). *Significant difference for participants versus nonparticipants, controlling for child’s age (p=.02). *No significant difference for nonparticipants versus control group.

Impact of Credit with Education !!!34 found between the nonparticipant and con- groups. In addition, the age at which chil- trol groups. dren in the three groups were weaned, al- These results are particularly encouraging, though not significantly different, was old- since participants might be expected to be est for participants (19.1 months) as com- using feeding bottles more than nonpartici- pared to nonparticipants (16.3 months) and pants. A necessary concern about Credit with controls (16.7 months). Education is that time-intensive child-care The predominant loan activities of women practices like breastfeeding might be compro- in the program—preparing and selling cooked mised as women invest more capital, and food, making and selling oil, and buying and potentially more of their own labor, into their selling foodstuffs—are not mutually exclusive loan. However, these results indicate that with child care that includes breastfeeding. participants were not more likely to use feed- Virtually all of the processing work to make ing bottles, and in fact after at least one year oil and prepare food is done at home with in the Credit with Education program, were less children nearby. The great majority of likely to use feeding bottles than women not women also sell their products in local, weekly in the program. markets, which requires minimal travel, and young children can typically accompany their Breastfeed for Two Years mothers. For similar reasons, participants might also Breastfeeding Score be expected to wean their children earlier than nonparticipants, particularly if the de- Participation in the Credit with Education pro- mands of women’s loan activities required gram seems to have improved participants’ them to depend upon surrogate child care. breastfeeding practices. Between the baseline However, the results again showed that this and the follow-up period, Credit with Educa- type of negative trade-off was not occurring tion participants were significantly more likely as a result of program participation. In the to (1) give their newborns colostrum, (2) ex- follow-up round, participants’ children who clusively breastfeed longer, and (3) not use were 12 to 24 months of age were slightly feeding bottles. While no statistically signifi- more likely to still be breastfed (77%) as com- cant impact on duration of breastfeeding was pared to the children of nonparticipants found, it is encouraging that involvement in (69%) and of residents in control communi- loan activities does not seem to undermine ties (70%). As indicated in Figure 5.5, there women’s ability to breastfeed their children. was no significant difference among the three

Figure 5.5: Children (12-24 Months) Still Breastfeeding

100% 77% 80% 73% 70% 69% 71% 70%

60%

40%

20% % Respondents

0% Program Communities Program Communities Control Communities 123 Future Future 1993 1996 Participants Participants Nonpartic. Nonpartic. (n=99) (n=103) 1993 1996 1993 1996 (n=48) (n=90) (n=151) (n=109)

Comparing 1993 and 1996 *No significant difference for participants versus control group, controlling for child’s age. *No significant difference for participants versus nonparticipants, controlling for child’s age. *No significant difference for nonparticipants versus control group, controlling for child’s age.

Research Paper No. 4!!!35 Table 5.6: Age Watery Foods Like Porridges First Introduced Residents in Control Participants Nonparticipants Communities 1993 1996 1993 1996 1993 1996 N=45 N=44 N=134 N=103 N=87 N=90 Less than 5 96% 50% 86% 68% 93% 72% months 5 - 6 months 4% 50% 9% 24% 5% 19% 7 months and ------7% 5% 8% 2% 9% older

An overall breastfeeding behavior score was Education program. Comparison of the per- derived on the basis of the five breastfeeding centage of mothers introducing food at about ideal behaviors described above. Respon- six months shows a significant difference be- dents received a maximum score of five—one tween the years when controlling for the point each for (1) giving colostrum to new- child’s age for participants as compared to borns, (2) not giving water until a newborn nonparticipants (p=.05) although not as was at least 150 days old, (3) not giving compared to residents in control communi- watery foods until at least 150 days, (4) ties. never using a feeding bottle, and (5) if the To better understand whether program par- child was either still breastfeeding or was ticipation had affected the quality of comple- weaned after 24 months of age. In all three mentary foods, mothers were asked what groups, breastfeeding practices in general foods they had given their children when improved, although least of all for residents they were 6 to 9 months of age. Follow-up in the control communities. However, the questions specifically probed whether moth- greatest difference was evident for the par- ers had added enrichment to the maize por- ticipant group which had a baseline mean ridge, which is the most common and fre- breastfeeding score of 1.5 and a follow-up quently given complementary food. Table mean score of 3.4. Nonparticipants went 5.7 summarizes the frequency with which from a mean score of 1.6 to 2.7, and resi- various enrichments were reportedly added, dents in control communities went from 1.6 either during or after the preparation of the to 2.1. The difference between years for the porridge. (Children who were older than six overall breastfeeding score was statistically months when their mothers joined Credit positive and significant for the participants with Education were excluded from the par- relative to the other two groups. The differ- ticipant sample for the follow-up time pe- ence is even greater when the participant riod.) Although sugar was a common en- sample is restricted to those children born richment mentioned by women in each of after their mothers joined the program. the three groups, there was a relatively dra- Complementary Food matic increase of adding groundnut (paste or flour) among participants. In this topic area, the timing of introduction Controlling for the child’s age, there was a and quality of complementary foods was com- significant and positive difference between pared across the three impact evaluation years in the mean number of nutritional en- groups. Table 5.6 shows that, for the baseline richments that participants reported add- period across the three groups, foods like por- ing to koko relative to residents in control ridges were first given to children earlier than the recommendation of about 6 months of communities (p=.03). Respondents were age (defined as from 150-209 days). given one point for each of the following nutritional enrichments promoted by the pro- For this analysis, the 1996 participant gram: beans/cowpeas, egg, fish, groundnut, sample was limited to those children born milk and palm oil. In the 1996 follow-up after their mothers had joined the Credit with

Impact of Credit with Education !36 Table 5.7: Enrichment Added to Porridge Given to Children at 6 to 9 Months of Age Residents in Control Participants Nonparticipants Communities 1993 1996 1993 1996 1993 1996 N=36 N=44 N=114 N=75 N=77 N=72 Bean/Cowpea 6% 34% 3% 25% 3% 10% Egg 22% 34% 25% 28% 26% 22% Fish 8% 14% 7% 8% 5% 4% Groundnut 28% 61% 25% 39% 26% 31% Milk/Formula/ 47% 36% 33% 32% 30% 40% Weanimix Sugar 56% 59% 66% 69% 66% 65% Mean 1.1 1.8 .9 1.3 .9 1.1 nutritional enrichments period, participants reported adding almost health centers and at monthly growth-moni- two nutritional enrichments as compared to toring sessions held in many of the program only one for the controls. However, it seems area communities. It is typically sold in small program promotion for adding legumes plastic bags in amounts that would be suffi- (groundnut paste or flour and bean or cow- cient for several servings only. pea flour) has been more successful than the Learning sessions facilitated at the Credit As- message to add fish powder. Only a third of sociation meetings encouraged mothers to the participants reported enriching koko with give Weanimix to their babies who were six fish powder. Given the benefits of animal months and older. Mothers were also en- proteins to iron absorption, resistance and couraged to add fish powder to the Weanimix obstacles to this practice should be explored. porridge to make it a particularly good Through its Maternal and Child Health complementary food. The baseline and fol- (MCH) services, the Ministry of Health in low-up surveys included questions to assess Ghana promotes a complementary food for mothers’ knowledge and trial of Weanimix. babies six months and older that is called Table 5.8 shows that even during the baseline “Weanimix.” Weanimix is a porridge that period, the majority of mothers had heard combines one-part maize flour to ½-parts of Weanimix. Although the percentage of bean and groundnut flour. To encourage its 1996 participants who had heard of adoption, Weanimix flour is prepared by the Weanimix was higher than the other two MCH program and sold both at the local groups, there was no significant difference in

Table 5.8: Knowledge and Trial of Weanimix Residents in Control Participants Nonparticipants Communities 1993 1996 1993 1996 1993 1996 N=48 N=66 N=150 N=111 N=99 N=102 Had heard of 77% 94% 79% 87% 77% 86% Weanimix Had tried 48% 62% 51% 57% 48% 44% Weanimix If tried ----- 68% ----- 40% ----- 41% Weanimix, added fish powder

Research Paper No. 4!37 the changes between years. When control- gram.) For those mothers who had tried ling for the child’s age, there was a margin- Weanimix, and controlling for the child’s age, ally significant difference in change between participants were significantly more likely to years for participants as compared to resi- enrich the Weanimix with fish powder than dents in control communities (p=.06) in were nonparticipants (p=.006) or residents whether mothers reported trying Weanimix. in control communities (p=.001). However, there was no significant difference between participants and nonparticipants. The impact survey also included questions (Again, the 1996 participant sample was re- to assess the current status of the study stricted to include only those children who children’s diets. To evaluate feeding fre- were six months or younger when their quency, mothers were asked to report how mother joined the Credit with Education pro-

Child Feeding In addition to breastfeeding, the program also promotes ideal behaviors about how, when, what and how often young children should be fed to promote their healthy growth. Learning sessions address the following:

Topic Specific message or ideal behavior • When to introduce • when babies are about 6 months of age complementary foods • Appropriate complementary • enriched porridges such as Weanimix foods (porridge made from maize and legume flour) enriched with fish powder • enriched traditional porridges like koko (maize-based porridge) made thicker than usual and enriched with either groundnuts, beans, fish powder, egg or milk • mpotompoto (mashed yam or cocoyam) enriched with fish powder and palm oil • nutritious snacks like mashed fruits and mashed vegetables • Increase feeding frequency • in addition to breastmilk, children 8 to 24 months should be fed a meal or nutritious