You are here

Article Text

Article menu

Download PDFPDF

Other topics
Microcredit participation and child health: results from a cross-sectional study in Peru
  1. H Moseson1,
  2. R Hamad2,
  3. L Fernald3
  1. 1Department of Epidemiology & Biostatistics, University of California, San Francisco, California, USA
  2. 2Division of General Medical Disciplines, Stanford University, Stanford, California, USA
  3. 3Division of Community Health and Human Development, School of Public Health, University of California, Berkeley, California, USA
  1. Correspondence to Heidi Moseson, Department of Epidemiology & Biostatistics, University of California, San Francisco, 185 Berry Street, Suite 5700, San Francisco, CA 94158, USA; heidi.moseson{at}ucsf.edu

Abstract

Background Childhood malnutrition is a major consequence of poverty worldwide. Microcredit programmes—which offer small loans, financial literacy and social support to low-income individuals—are increasingly promoted as a way to improve the health of clients and their families. This study evaluates the hypothesis that longer participation in a microcredit programme is associated with improvements in the health of children of microcredit clients.

Methods Cross-sectional data were collected in February 2007 from 511 clients of a microcredit organisation in Peru and 596 of their children under 5 years of age. The primary predictor variable was length of participation in the microcredit programme. Outcome variables included height, weight, anaemia, household food security and parent-reported indicators of child health. Multivariate linear and logistic regressions assessed the association between the number of loan cycles and child health outcomes. Pathways through which microcredit may have influenced health outcomes were also explored via mediation analyses.

Results Longer participation in microcredit was associated with greater household food security and reduced likelihood of childhood anaemia. No significant associations were observed between microcredit participation and incidence of childhood illnesses or anthropometric indicators. Increased consumption of red meat may mediate the association between the number of loan cycles and food security, but not the association with anaemia.

Conclusions The effects of microcredit on the health of clients’ children are understudied. Exploratory findings from this analysis suggest that microcredit may positively influence child health, and that diet may play a causal role.

  • Economic evaluation
  • NUTRITION
  • CHILD HEALTH
  • EPIDEMIOLOGY
  • INTERNATIONAL HLTH

https://doi.org/10.1136/jech-2014-204071

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Introduction

    Childhood malnutrition is one of the most harmful health consequences of poverty and a great threat to public health worldwide.1 Children throughout low and middle-income countries (LMIC) suffer the consequences of malnutrition, which include death, impaired growth and impaired cognitive, language and motor skills.2 ,3 A variety of biomedical interventions to address malnutrition and underdevelopment have been tested with differing degrees of success.4 Other approaches address the broader socioeconomic determinants of health disparities by focusing on poverty alleviation and reductions in inequality.5

    Microcredit programmes are one type of poverty alleviation intervention, offering small loans to low-income individuals who otherwise lack access to financial services.6 The hope is that clients (often women) invest these loans in their businesses to improve their families’ health and living conditions. Women often meet in groups, providing each other with social collateral to make up for their lack of physical collateral. Organisations that offer microcredit lending are often non-profits with a social mission; many integrate the loans with other programmes to promote financial literacy, women's empowerment or health.

    Microcredit is often promoted as an intervention to improve socioeconomic status and health.7 ,8 A review of the literature reveals that most studies focus on the health of clients themselves rather than their children. Microcredit participation has been associated with higher haemoglobin levels and food security,9 ,10 and increased use of sanitary latrines and contraception,11 yet not all results have been positive. For example, studies have suggested both increased and decreased healthcare utilisation among microcredit clients,12 ,13 and inconsistent effects on mental health.14–16 Given that many communities are saturated with microcredit lenders,17 randomised controlled trials are difficult to conduct: loan take-up is often higher in the control group, which may lead to an underestimate of the effects of the loans.18 One study to employ such a design involved the randomisation of loans to slums in India, with results demonstrating no impact on health, sanitation or food expenditures.18 Further, the structure of microcredit interventions varies significantly across programmes, making it difficult to make generalisations about programme impact. Several review articles on the effects of microcredit programmes on economic, health and social outcomes find that the quantitative evidence that exists is sparse and inconclusive.19–22

    The few studies on health outcomes among the children of microcredit clients suffer from similar limitations. One study found that children of clients who had access to loans had greater height-for-age than children of clients with limited access,10 while an evaluation of a microcredit programme in Bolivia with an educational intervention found that participation was associated with improvements in weight-for-age among programme children, compared with children of non-participants.23 A credit-with-education programme in Ghana resulted in higher height-for-age among 1-year-old children of participants compared with children of future participants.24 These last two studies did not distinguish whether the outcomes were a result of the microcredit loans or the educational intervention.

    Several pathways may link microcredit with child health. Increased income from loans may lead to greater spending on health, food, sanitation or housing.14 In particular, increased income may allow for purchase of higher priced foods with higher nutrient density, such as red meat, a food that is associated with higher intake of calories, iron, zinc and vitamin B12.25 Another pathway suggests that clients’ heightened empowerment as a result of participation in microcredit programmes leads them to adopt positive health behaviours.26 ,27 For example, women's increased control of their finances is believed to augment their decision-making power on issues such as household spending and contraceptive use. A third pathway proposes that the supplemental health education offered in conjunction with microcredit loans may lead to positive changes in client behaviour.28–30 A final pathway involves well-known mechanisms through which improvements in maternal health bring about improvements in child health31; a prior study of this sample found that microcredit participation was improved with reductions in maternal anaemia, such that this pathway may be important for children in this study.9 Conversely, since women participating in microcredit programmes typically continue to be responsible for housework in addition to their responsibilities in running a business and attending loan group meetings, this dual burden may reduce the time available for childcare or personal healthcare, leading to worsened family health outcomes.32 ,33

    Despite the paucity of rigorous studies on the association between microcredit participation and child health, over 3000 microcredit programmes exist around the world.34 This study aims to explore the impact of microcredit participation by examining the associations between child health and length of participation among clients of Prisma, a microcredit programme in Peru. Prisma provides loans through a group-lending model, with interest rates typical of those in other LMIC. Given the challenges discussed above in implementing randomised controlled studies of this type of intervention, our study relies on a cross-sectional design, aiming to test preliminary hypotheses about the association between microcredit and child health to inform future longitudinal research. We hypothesise that the additional household resources gained by participating in a microcredit programme for longer amounts of time, as well as improvements in parental health and empowerment, are associated with improved health outcomes among children of clients. The study adds to the existing literature by exploring potential pathways through which this poverty alleviation intervention may influence child health.

    Methods

    Study design

    Data were collected in 2007 among clients of Prisma, a non-profit microcredit organisation living in Pucallpa, Peru (population ∼136 000, 93% urban). The gross national income in Peru for 2007 was US$3330, and the Gini coefficient was 51.7, while 42% of the population was living below the national poverty line.35 At the time of this study, clients were not receiving any services from Prisma other than microcredit loans.

    All 2134 clients at the Pucallpa site were approached, and 1855 (87.7%) agreed to participate. The main reasons for non-response were refusal to participate and the client not being available for interview. We included in the analyses male and female clients who had children under the age of 5 years (n=511) and the children themselves (n=596). As Prisma had no data on clients’ children, it is unclear whether our sample is representative of other children of Prisma clients. The majority of adult participants in our sample were female (85%, n=436), and 48% of our sample children were girls (n=289). Given the sample size, our study is powered at the 80% level to detect small differences in the various health outcomes with an α of 5%, including in subgroup analyses.

    The questionnaire was translated into Peruvian Spanish by native speakers also fluent in English, and was piloted and validated. Local Spanish-speaking surveyors were trained by study staff. Clients were approached for interviews at their monthly loan group meetings. Those not attending were telephoned or approached at home. Clients were asked to bring their children under the age of 5 years to the loan group meeting for collection of anthropometric measurements. If a client did not bring his or her children to the meeting, interviewers travelled to the client's home for measurement. Surveyors used wooden stadiometers, constructed locally, to measure children's height. Weight was measured with digital Taylor Electric Lithium Scales model 7324W (Taylor Precision Products, Las Cruces, New Mexico). Haemoglobin levels were measured with finger sticks using the HemoCue Hb 201+ System (HemoCue Inc, Lake Forest, California, USA). Each client was also asked to report on a variety of health outcomes for each child under age 5.

    The Institutional Review Boards of the University of California Berkeley and Prisma provided ethics approval for this study. Clients provided written or verbal consent for participation on behalf of themselves and their children.

    Intervention

    Each month, loan groups of 10–20 clients meet with a loan officer from Prisma. During these meetings, clients repay their loans and address administrative matters. Prisma's average loan size is US$357 (range US$103–US$1199, SD US$209), which is received in a lump sum and repaid over the course of 6 months at a monthly interest rate of 4%. These rates are typical of microcredit organisations in the developing world.36 New members are approved by Prisma and organise themselves into groups, with each new member approved by existing members—a structure that allows poor individuals to provide social collateral for one another to make up for the lack of physical collateral.37

    Measures

    The primary independent variable was length of participation in Prisma's microcredit programme, as measured by the number of loan cycles completed by the client. The duration of each loan cycle is 6 months. This variable was transformed into a categorical variable to capture non-linearity: 0 loan cycles (reference group), 1 loan cycle, 2 loan cycles and 3–9 loan cycles. Clients with 0 loan cycles represent those individuals who are the newest members of Prisma, with less than 6 months of exposure to the programme. Those with 3–9 cycles (ie, 18–54 months) are the longest standing members. Clients who had completed 3–9 loan cycles were grouped together to reflect their shared, previously demonstrated ability to repeatedly repay loans. Using the duration of participation (measured in loan cycles) is a method that has been used in previous studies to capture the ‘dose–response’ association between exposure to the microcredit programme and health outcomes.9 ,14 Non-programme participants were explicitly not selected as a control group because they may differ from those individuals who choose to enrol in the programme in important ways related to motivation, entrepreneurship and other unmeasured factors.38

    The questionnaire included questions about demographics and socioeconomic status. Child age was included as a dichotomous variable (above and below 2 years of age). This was to capture potential non-linear effects, given that early childhood represents a critical developmental window during which environmental changes may have more of an effect on child health.39 Adult age was included as a continuous variable. Education was included as a categorical variable reflecting completion of primary school or less (reference group), some secondary education, complete secondary education and some post-secondary education. Clients were asked about household assets—such as refrigerators, cars and washing machines—characteristics of their home—such as construction materials—and whether their home had electricity. These measures have been shown to be valid indicators of poverty across regional and national samples.40 For each group of household indicators—assets, housing and electricity—a principal components variable was constructed as a measure of socioeconomic status. This method has been used in previous studies to measure relative poverty among households in developing countries.41 ,42 We modelled household socioeconomic status using household assets because of established methodological issues with accurate measurement of household income via self-report.43

    We next selected anthropometric measures that have been shown to be important markers of child poverty.44 These included length-for-age and weight-for-age Z-scores, calculated using reference standards of the WHO.45 The length of children under 2 years was measured while they were in the supine position. Blood haemoglobin levels were obtained using the HemoCue Hb 201+ System (HemoCue Inc). Anaemia was defined as a haemoglobin level under 11 mg/dL.46 Haemoglobin was not adjusted for altitude, as Pucallpa lies at only 200 m above sea level.46

    Clients were asked about their children's health in the last month, including the number of respiratory infections and diarrhoeal episodes. A nine-item index adapted from the US Household Food Security Survey Module was used to measure food insecurity, which captures the availability and affordability of food in the household in the past month, specifically focusing on insufficient quantity and periods of hunger. A score above 2 indicates household insecurity.47 This version has been validated previously in developing countries.48 We also asked parents about their children's dietary consumption during the past month using a Food Frequency Questionnaire (FFQ) that was modified to include common local foods. Consumption of red meat was measured by asking about the total number of times a child consumed red meat in the past month. We selected red meat for a food security mediation analysis because of its status as a higher cost luxury food, which is most likely more attainable as household income rises; we selected it for an anaemia mediation analysis because it is iron-rich, and iron is known to affect haemoglobin levels.49 We also asked about household water sources and whether water was properly treated, defined as having been commercially bottled, boiled or treated with chlorine or lye. For anthropometric measures and survey questions, validity testing was conducted prior to survey launch to ensure inter-rater and intra-rater reliability.

    Some outcome measurements are missing (0–8% missing, depending on the outcome), seemingly randomly, due to parental refusal, insufficient time or surveyor error. We include only those children with data on relevant outcomes.

    Data analysis

    Survey staff double-entered data using CSPro 3.3 (US Census Bureau, Population Division, Washington DC, USA). Analyses were conducted using Stata IC 13 (College Station, Texas, USA).

    We used a cross-sectional design to investigate the association between clients’ length of participation in the microcredit programme and health outcomes among their children. Mean values and frequencies were calculated for client and child sociodemographic characteristics, as well as for health outcomes, overall and by number of loan cycles completed. Differences in mean values between loan groups were assessed with one-way analysis of variance or independent t tests. We used multivariate linear and logistic regressions to explore the association between loan cycles and child health while controlling for the influence of possible confounders. Regressions control for age and gender of the child and client, educational and marital status of the client, and household wealth, with errors clustered by household. We also constructed two interaction terms: number of loan cycles by child age and number of loan cycles by child gender. We were unable to explore differences by client gender because of the small number of male clients in our sample (15%). We consider both main terms and interaction terms with a p value less than 0.05 to be statistically significant.50 Hausman testing indicated that random effects models should be used. Regressions involving food insecurity were only conducted on first-born children for whom we have measures (n=493), as all children in the same household have the same food insecurity score.

    We also conducted a mediation analysis to measure the extent to which increased red meat consumption may mediate the relationship between loan cycles and improved food security, and between loan cycles and reduced anaemia. We used seemingly unrelated regression51 to compute indirect and direct effect coefficients for loan cycles and red meat in relation to food security score, and then calculated the proportion of the total effect of loan cycles on the food security score that was mediated (indirect effect) through red meat consumption.52 ,53 To account for the binary nature of anaemia and the logistic model used, we first rescaled the coefficients by the SDs of the predictor and outcome variables and then applied the product of coefficients approach.54 These calculations assume no unmeasured confounders of the mediator–outcome relationship, no causal intermediaries and no interaction between the exposure and mediator.55–57 The results from the binary mediation analysis for anaemia must be cautiously interpreted given the non-collapsibility property of the OR. We calculated 95% CIs for the mediation coefficients using bootstrap SEs, a non-parametric method based on resampling, with 500 repetitions using the bootstrap command in Stata V.13.

    Results

    Sample characteristics

    Seventy-five per cent of clients completed a secondary school education (table 1) and a mean of 2.2 loan cycles (range 0–9). Clients who had participated for a greater number of loan cycles were slightly older and had greater household assets (0.1 vs −0.3, p=0.01).

    View this table:
    Table 1

    Demographic characteristics of clients and their children, by loan cycles completed

    Children's ages ranged from 0 to 5 years (66% between 2 and 5 years), and 85% were first-born children (table 1). The prevalence of anaemia and average height-for and weight-for age in this sample is comparable to those in Peru more generally, suggesting chronic malnutrition.58 ,59 There were no meaningful differences in demographic characteristics of children across the number of loan cycles (table 2).

    View this table:
    Table 2

    Health outcomes of client children, by number of loan cycles completed

    Multivariate regressions

    In adjusted models (table 3), each additional loan cycle (6 months of programme exposure) was associated with 11% lower odds of anaemia as compared with children of clients who had completed one fewer loan cycle (OR=0.89, 95% CI 0.80 to 1.01, p=0.06). We found no evidence that either child age or gender modified the relationship between the number of loan cycles and anaemia.

    View this table:
    Table 3

    Health outcomes among children of microcredit participants with number of loan cycles as predictor

    Each additional loan cycle was associated with a decrease of 15% in the food insecurity score (β=−0.15, 95% CI −0.26 to −0.04) in adjusted models, implying an improvement in food security. We found no evidence for effect measure modification by child age or gender. We also found no evidence of an association between length of participation in a microcredit programme and child height, weight or recent illness.

    Completion of more loan cycles was borderline associated with an increase in a child's mean red meat consumption, with the most striking difference between children whose families have completed one loan cycle versus none (p=0.06; table 4). Children in food-secure households consumed significantly more red meat servings per day on average than did those in food-insecure households (mean difference=0.06, p=0.04). A mediation analysis quantified this relationship, identifying that approximately 9% (95% CI 0% to 25%) of the association between the number of loan cycles completed and household food insecurity is explained by children's red meat consumption. We observed no difference in average red meat consumption between anaemic and non-anaemic children (table 4), and a mediation analysis confirmed that red meat consumption did not mediate the relationship between loan cycles and anaemia (95%CI −0.08 to 0.05). No difference in access to clean water was seen across loan groups or by anaemic status.

    View this table:
    Table 4

    Analysis of pathways through which microcredit loans may influence health

    Discussion

    We hypothesised that longer participation in a microcredit programme would be associated with better health outcomes among clients’ children. We found that longer participation was associated with a marginally significantly lower likelihood of anaemia among children and a significantly greater household food security; the food security findings appear to operate partially via an increase in red meat consumption. We found no evidence for an association between the number of loan cycles completed and child height, weight or incidence of other illnesses, and no difference in outcomes by child age or gender.

    Other studies have also found that participation in a microcredit programme is associated with improvements in household food security,10 ,23 although they did not identify factors that contributed to this improvement. We explored a possible pathway by which microcredit may improve household food security, through the increased consumption of red meat. As the food security questionnaire is a subjective measure of the respondent's perception of food access and availability, and as red meat consumption often symbolises wealth in many cultures,60 a respondent may report improved perceptions of food security based on increased consumption of red meat.

    Our finding of improved anaemia levels among clients’ children is similar to a finding in a separate study of women in our sample, in which we found that haemoglobin levels were higher among female clients with longer participation in the programme.9 While we do not find that this is due to the increase in red meat consumption, it may be that the children had greater access to other categories of iron-rich foods not addressed in this study, such as leafy green vegetables or legumes. We also did not find that this relationship was mediated by access to clean water that may have improved health status by reducing gastrointestinal illnesses. It is also possible that higher incomes lead to greater household investments in malaria treatment and prevention, which reduce the burden of anaemia, although this is not a pathway we examined. Intra-household distribution of food may also have been altered given the improved food security,61 such that children may have received a larger share of nutritional resources with a subsequent improvement in anaemia.

    Consistent with the few previous studies, our study finds no association between programme participation and other anthropometric or patient-reported health outcomes in children.23 ,24 ,62 This may be because participation in microcredit leads to a different, but not necessarily better distribution of household resources. A household may feel more able to buy more food and thus improve food security, but these dietary changes may not be sufficient to change anthropometric measures. Our analyses of access to clean water seem to confirm this: longer participation in the microcredit programme is not associated with improvements in the water supply that could have reduced diarrhoeal illnesses. This suggests that public health practitioners and socially conscious microcredit organisations may need to look beyond a ‘food-first’ bias, instead focusing on multi-sectoral infrastructural changes to address malnutrition and child health.63

    Our study has significant implications for public health activities to improve child health and nutrition. Health researchers have long acknowledged that addressing social and economic factors is critical in alleviating health disparities,5 and some have suggested that microcredit may be an appropriate tool to address these factors in order to attain the Millennium Development Goals.6 Even though economic interventions such as microcredit target the more distal factors rather than those more proximal to a child's immediate environment, our study supports the idea that they should be integrated into other public health interventions to achieve these aims.

    There are several limitations to this study. The use of a cross-sectional design leaves us unable to establish causality (or to reject reverse causality) or track individual changes over time. Further, our use of new members as the control group raises several issues. New clients may fundamentally differ from long-term clients in terms of stage of family life and eagerness to take up the intervention, depending on how the programme was rolled out. Thus, our results may be biased according to these unmeasured differences. One major strength of this approach, however, is that both our exposed and control groups are comparable with respect to the qualities that lead a person to join a microcredit organisation. Another limitation involves the use of a 1-month recall for dietary intake, which may lead to measurement error in the FFQ due to recall bias. A final limitation is that microcredit programmes are known to differ substantially from one another in terms of structure and content. These differences complicate efforts to make valid comparisons across programmes, which limits the generalisability of our findings to those programmes with similar structure. Moreover, given that microcredit organisations differ in terms of the social contexts in which they operate, the results should be applied with caution to programmes operating in other international settings. For example, added income may not be as beneficial for child health and nutrition in a community in which the individuals still lack access to healthcare, nutritious food and other resources.

    Conclusions

    In this study among children of microcredit participants in Peru, we evaluate an intervention that may improve child health by addressing poverty, one of the leading socioeconomic determinants of health. Our results suggest that participation in this microcredit programme is associated with greater household food security and lower levels of anaemia among clients’ children. We found that increased consumption of red meat may partially explain this finding. Further research should attempt to elucidate more exactly the mechanisms through which microcredit participation may influence these child health outcomes, and how and when we can intervene to make sure that similar economic interventions effectively improve the health and well-being of children.

    What is already known on this subject

    • Only a handful of studies have explored the influence of microcredit participation on the health of clients’ children, and they suggest that there may be a positive link between the two. No studies, however, have quantitatively measured which mechanisms of microcredit participation may influence the health of clients’ children.

    What does this study add

    • This study finds that longer participation in a microcredit programme is associated with improved household food security and reduced childhood anaemia. Specifically, findings suggest that dietary changes, enabled by microcredit loans, may explain these beneficial health outcomes in children of microcredit clients in Peru.

    Acknowledgments

    The authors would like to thank Dean Karlan, Maureen Lahiff, Silvia Robles, Miguel Almunia and Tania Alfonso for their assistance with this study, as well as Prisma, the surveyors and the study subjects for their participation. They also acknowledge the reviewers for their valuable contributions.

    References

    1. ↵
      1. Blössner M,
      2. De Onis M,
      3. Prüss-Üstün A,
      4. et al
      . Malnutrition: quantifying the health impact at national and local levels. 2005. Geneva, World Health Organization. WHO Environmental Burden of Disease Series, No. 12.
    2. ↵
      1. Caulfield LE,
      2. de Onis M,
      3. Blössner M,
      4. et al
      . Undernutrition as an underlying cause of child deaths associated with diarrhea, pneumonia, malaria, and measles. Am J Clin Nutr 2004;80:193–8.
      OpenUrlAbstract/FREE Full Text
    3. ↵
      1. Walker SP,
      2. Wachs TD,
      3. Grantham-McGregor S,
      4. et al
      . Inequality in early childhood: risk and protective factors for early child development. Lancet 2011;378:1325–38.
      OpenUrlCrossRefPubMedWeb of Science
    4. ↵
      1. Bhutta ZA,
      2. Das JK,
      3. Rizvi A,
      4. et al
      . Evidence-based interventions for improvement of maternal and child nutrition: what can be done and at what cost? Lancet 2013;382:452–77.
      OpenUrlCrossRefPubMedWeb of Science
    5. ↵
      1. Marmot M,
      2. Friel S,
      3. Bell R,
      4. et al
      . Closing the gap in a generation: health equity through action on the social determinants of health. Lancet 2008;372:1661–9.
      OpenUrlCrossRefPubMedWeb of Science
    6. ↵
      1. Littlefield E,
      2. Morduch J,
      3. Hashemi S
      . Is microfinance an effective strategy to reach the Millennium Development Goals? Focus Note 2003;24:1–11.
      OpenUrl
    7. ↵
      1. Leatherman S,
      2. Metcalfe M,
      3. Geissler K,
      4. et al
      . Integrating microfinance and health strategies: examining the evidence to inform policy and practice. Health Policy Plan 2012;27:85–101.
      OpenUrlAbstract/FREE Full Text
    8. ↵
      1. Dunford C,
      2. MkNelly B
      . Using microfinance to improve health and nutrition security. Global HealthLink 2002;118:9, 22.
      OpenUrl
    9. ↵
      1. Hamad R,
      2. Fernald LC
      . Microcredit participation and nutrition outcomes among women in Peru. J Epidemiol Community Health 2012;66:e1.
      OpenUrlAbstract/FREE Full Text
    10. ↵
      1. Guha-Khasnobis B,
      2. Hazarika G
      . Household access to microcredit and children's food security in rural Malawi: a gender perspective. Research Paper, UNU-WIDER, United Nations University (UNU), 2007.
    11. ↵
      Bangladesh Institute of Development Studies. Final Report on BIDS Study on PKSF's Monitoring and Evaluation System. Dhaka, Bangladesh, 2001.
    12. ↵
      1. Ahmed SM,
      2. Petzold M,
      3. Kabir ZN,
      4. et al
      . Targeted intervention for the ultra poor in rural Bangladesh: does it make any difference in their health-seeking behaviour? Soc Sci Med 2006;63:2899–911.
      OpenUrlCrossRefPubMed
    13. ↵
      1. Coleman BE
      . The impact of group lending in Northeast Thailand. J Dev Econ 1999;60:105–41.
      OpenUrlCrossRefWeb of Science
    14. ↵
      1. Mohindra KS,
      2. Haddad S,
      3. Narayana D
      . Can microcredit help improve the health of poor women? Some findings from a cross-sectional study in Kerala, India. Int J Equity Health 2008;7:2.
    15. ↵
      1. Fernald LC,
      2. Hamad R,
      3. Karlan D,
      4. et al
      . Small individual loans and mental health: a randomized controlled trial among South African adults. BMC Public Health 2008;8:409.
      OpenUrlCrossRefPubMed
    16. ↵
      1. Ahmed S,
      2. Chowdhury M,
      3. Bhulya A
      . Micro-credit and emotional well-being: experience of poor rural women from Matlab, Bangladesh. World Dev 2001;29:1957–66.
      OpenUrlCrossRefWeb of Science
    17. ↵
      1. Krauss A,
      2. Lontzek L,
      3. Meyer J
      . Does market saturation increase the risk of over-indebtedness? Washington DC: Consultative Group to Assist the Poor, 2013.
    18. ↵
      1. Banerjee A,
      2. Duflo E,
      3. Glennerster R,
      4. et al
      . The miracle of microfinance? Evidence from a randomized evaluation. 2013. Working Paper. MIT Department of Economics, National Bureau of Economic Research, Abdul Latif Jameel Poverty Action Lab.
    19. ↵
      1. Goldberg N
      . Measuring the impact of microfinance: taking stock of what we know. Washington DC: Grameen Foundation USA, 2005.
    20. ↵
      1. Duvendack M,
      2. Palmer-Jones R,
      3. Copestake JG,
      4. et al
      . What is the evidence of the impact of microfinance on the well-being of poor people? London: EPPI-Centre, Social Science Research Unit, Institute of Education, University of London, 2011.
    21. ↵
      1. Odell K
      . Measuring the impact of microfinance. Washington: Grameen Foundation, 2010:1–38.
    22. ↵
      1. Orso CE
      . Microcredit and poverty. An overview of the principal statistical methods used to measure the programme net impacts; 2011 February, POLIS Working Paper No 180.
    23. ↵
      1. MkNelly B,
      2. Dunford C
      . Impact of credit with education on mothers and their young children's nutrition: CRECER credit with education program in Bolivia. Davis, CA: Freedom From Hunger, 1999.
    24. ↵
      1. MkNelly B,
      2. Dunford C
      . Impact of credit with education on mothers and their young children's nutrition: lower Pra rural bank credit with education program in Ghana. Davis, CA: Freedom From Hunger, 1998.
    25. ↵
      1. Prynee C,
      2. Wagemakers J,
      3. Stephen A,
      4. et al
      . Meat consumption after disaggregation of meat dishes in a cohort of British adults in 1989 and 1999 in relation to diet quality. Eur J Clin Nutr 2009;63:660–6.
      OpenUrlCrossRefPubMedWeb of Science
    26. ↵
      1. Ashburn K,
      2. Kerrigan D,
      3. Sweat M
      . Micro-credit, women's groups, control of own money: HIV-related negotiation among partnered Dominican women. AIDS Behav 2008;12:396–403.
      OpenUrlCrossRefPubMedWeb of Science
    27. ↵
      1. Schuler SR,
      2. Hashemi SM
      . Credit programs, women's empowerment, and contraceptive use in rural Bangladesh. Stud Fam Plann 1994;25:65–76.
      OpenUrlCrossRefPubMedWeb of Science
    28. ↵
      1. Pronyk PM,
      2. Hargreaves JR,
      3. Morduch J
      . Microfinance programs and better health: prospects for sub-Saharan Africa. JAMA 2007;298:1925–7.
      OpenUrlCrossRefPubMedWeb of Science
    29. ↵
      1. Hamad R,
      2. Fernald LC,
      3. Karlan DS
      . Health education for microcredit clients in Peru: a randomized controlled trial. BMC Public Health 2011;11:51.
      OpenUrlCrossRefPubMed
    30. ↵
      1. Smith SC
      . Village banking and maternal and child health: evidence from Ecuador and Honduras. World Dev 2002;30:707–23.
      OpenUrlCrossRefWeb of Science
    31. ↵
      1. Black RE,
      2. Allen LH,
      3. Bhutta ZA,
      4. et al
      . Maternal and child undernutrition: global and regional exposures and health consequences. Lancet 0000;371:243–60.
      OpenUrlWeb of Science
    32. ↵
      1. Brett J
      . “We sacrifice and eat less”: the structural complexities of microfinance participation. Hum Organ 2006;65:8–19.
      OpenUrlWeb of Science
    33. ↵
      1. Mayoux L
      . Women's empowerment and micro-finance programmes: strategies for increasing impact. Dev Pract 1998;8:235–41.
      OpenUrlCrossRefPubMed
    34. ↵
      1. Reed L
      . State of the microcredit summit campaign report. Washington DC: Microcredit Summit Campaign, 2011.
    35. ↵
      World Bank. Data by Country: Peru.
    36. ↵
      1. Easton T
      . The hidden wealth of the poor: a survey of microfinance. Economist 2005;377. Special Report: Microfinance. http://www.economist.com/node/5079324
    37. ↵
      1. Yunus M
      . Toward eliminating poverty from the world: Grameen Bank experience. In: Anderson CL, Looney JW, eds. Making progress: essays in progress and public policy. Lanham, MD: Lexington Books, 2002:371–8.
    38. ↵
      1. Mosley P
      . The use of control groups in impact assessment for microfinance. Department of Economics and Department of Agricultural Economics, University of Reading, 1997.
    39. ↵
      1. Bornstein MH
      . Sensitive periods in developments: interdisciplinary perspectives. Hillsdale, NJ: Lawrence Erlbaum Publishers, 1987.
    40. ↵
      MEASURE DHS. Demographic and Health Surveys. Calverton, MD 2009.
    41. ↵
      1. Filmer D,
      2. Pritchett L
      . Estimating wealth effects without expenditure data—or tears: an application to educational enrollments in states of India. Policy Research Working Papers No. 1994. Washington DC: World Bank, 1998.
    42. ↵
      1. Zeller M,
      2. Houssou N,
      3. Alcaraz GV,
      4. et al
      . Developing poverty assessment tools based on principal component analysis: results from Bangladesh, Kazakhstan, Uganda, and Peru. International Association of Agricultural Economists 2006 Annual Meeting. Queensland, Australia 2006.
    43. ↵
      1. Moore JC,
      2. Stinson LL,
      3. Welniak EJ
      . Income measurement error in surveys: a review. J Official Stat 2000;16:331–61.
      OpenUrl
    44. ↵
      1. Brooks-Gunn J,
      2. Duncan GJ
      . The effects of poverty on children. Future Child 1997;7:55–71.
      OpenUrlCrossRefPubMedWeb of Science
    45. ↵
      WHO. Physical status: the use and interpretation of anthropometry. Report of a WHO expert committee. Technical report series no. 854. Geneva 1995:854.
    46. ↵
      WHO. Iron deficiency anaemia: assessment, prevention, and control. A Guide for Programme Managers. Geneva 2001. A publication of the World Health Organization.
    47. ↵
      Service UER. Food security in the US: survey tools. USDA 2013.
    48. ↵
      1. Melgar-Quinonez HR,
      2. Zubieta A,
      3. MkNelly B,
      4. et al
      . Household food insecurity and food expenditure in Bolivia, Burkina Faso, and the Philippines. J Nutr 2006;136:1431S–7S.
      OpenUrlAbstract/FREE Full Text
    49. ↵
      1. Olaya GA,
      2. Lawson M,
      3. Fewtrell MS
      . Efficacy and safety of new complementary feeding guidelines with an emphasis on red meat consumption: a randomized trial in Bogota, Colombia. Am J Clin Nutr 2013;98:983–93.
      OpenUrlAbstract/FREE Full Text
    50. ↵
      1. Marshall SW
      . Power for tests of interaction: effect of raising the type I error rate. Epidemiol Perspect Innov 2007;4:4.
      OpenUrlCrossRefPubMed
    51. ↵
      1. Zellner A
      . An efficient method of estimating seemingly unrelated regressions and tests for aggregation bias. J Am Stat Assoc 1962;57:348–68.
      OpenUrlCrossRefWeb of Science
    52. ↵
      1. Baron RM,
      2. Kenny DA
      . The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. J Pers Soc Psychol 1986;51:1173.
      OpenUrlCrossRefPubMedWeb of Science
    53. ↵
      1. Preacher KJ,
      2. Hayes AF
      . Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behav Res Methods 2008;40:879–91.
      OpenUrlCrossRefPubMedWeb of Science
    54. ↵
      1. MacKinnon DP,
      2. Dwyer JH
      . Estimating mediated effects in prevention studies. Eval Rev 1993;17:144–58.
      OpenUrlAbstract/FREE Full Text
    55. ↵
      1. Cole S,
      2. Hernán M
      . Fallibility in estimating direct effects. Int J Epidemiol 2002;31:163–5.
      OpenUrlAbstract/FREE Full Text
    56. ↵
      1. Petersen M,
      2. Sinisi S,
      3. van der Laan M
      . Estimation of direct causal effects. Epidemiology 2006;17:276–84.
      OpenUrlCrossRefPubMedWeb of Science
    57. ↵
      1. VanderWeele T,
      2. Vansteelandt S
      . Odds ratios for mediation analysis for dichotomous outcome. Am J Epidemiol 2010;172:1339–48.
      OpenUrlAbstract/FREE Full Text
    58. ↵
      1. Benoist Bd,
      2. McLean E,
      3. Egll I,
      4. et al
      . Worldwide prevalence of anaemia 1993–2005: WHO global database on anaemia. World Health Organization, 2008.
    59. ↵
      1. Castro Bedriñana J,
      2. Chirinos Peinado D
      . Z-score anthropometric indicators derived from NCHS-1977, CDC-2000 and WHO-2006 in children under 5 years in Central Area of Peru. Universal J Public Health 2014;2:73–81.
      OpenUrl
    60. ↵
      1. Allen MW,
      2. Baines S
      . Manipulating the symbolic meaning of meat to encourage greater acceptance of fruits and vegetables and less proclivity for red and white meat. Appetite 2002;38:118–30.
      OpenUrlCrossRefPubMedWeb of Science
    61. ↵
      1. Messer E
      . Intra-household allocation of food and health care: current findings and understandings—Introduction. Soc Sci Med 1997;44:1675–84.
      OpenUrlCrossRefPubMedWeb of Science
    62. ↵
      1. Pitt MM,
      2. Khandker SR
      . The impact of group-based credit programs on poor households in Bangladesh: does the gender of participants matter? J Political Economy 1998;106:958–96.
      OpenUrlCrossRefWeb of Science
    63. ↵
      1. Pelletier DL,
      2. Deneke K,
      3. Kidane Y,
      4. et al
      . The food-first bias and nutrition policy: lessons from Ethiopia. Food Policy 1995;20:279–98.
      OpenUrlCrossRefPubMedWeb of Science
    View Abstract

    Footnotes

    • Contributors HM and RH developed the research question and analysis plan, with input from LF. HM conducted all analyses and drafted the paper, with guidance and input from RH and LF at each stage. All authors reviewed and approved the final draft.

    • Funding This project was supported financially by the American Women's Hospitals Services, the Bixby Program at the University of California Berkeley (UCB), the Center for Latin American Studies at UCB, the Dean's Summer Fellowship at the University of California San Francisco (UCSF), the Human Rights Center at UCB, Innovations for Poverty Action, the Interdisciplinary MPH Program at the UCB School of Public Health, the Rainer Fund, the UCSF-UCB Joint Medical Program, and the National Institute for General Medical Sciences (NIGMS) Initiative for Maximizing Student Development Fellowship Grant number R25 GM56847.

    • Competing interests None.

    • Patient consent Obtained.

    • Ethics approval University of California, Berkeley IRB.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Data sharing statement De-identified from the study may be made available to interested parties upon review of any requests, as determined by the study PIs.