Is microfinance associated with changes in women’s well-being and children’s nutrition? A systematic review and meta-analysis

Background Microfinance is the provision of savings and small loans services, with no physical collateral. Most recipients are disadvantaged women. The social and health impacts of microfinance have not been comprehensively evaluated. Objective To explore the impact of microfinance on contraceptive use, female empowerment and children’s nutrition in South Asia, Sub-Saharan Africa and Latin America and the Caribbean. Design We conducted a systematic search of published and grey literature (1990–2018), with no language restrictions. We conducted meta-analysis, where possible, to calculate pooled ORs. Where studies could not be combined, we described these qualitatively. Data sources EMBASE, MEDLINE, LILACS, CENTRAL and ECONLIT were searched (1990–June 2018). Eligibility criteria We included controlled trials, observational studies and panel data analyses investigating microfinance involving women and children. Data extraction and synthesis Two independent reviewers extracted data and assessed risk of bias. The methodological quality of included studies was assessed using the Cochrane risk-of-bias tool for controlled trials and quasi-experimental studies and a modified Newcastle Ottawa Scale for cross-sectional surveys and analyses of panel data. Meta-analyses were conducted using STATA V.15 (StataCorp). Results We included 27 studies. Microfinance was associated with a 64% increase in the number of women using contraceptives (OR 1.64, 95% CI 1.45 to 1.86). We found mixed results for the association between microfinance and intimate partner violence. Some positive changes were noted in female empowerment. Improvements in children’s nutrition were noted in three studies. Conclusion Microfinance has the potential to generate changes in contraceptive use, female empowerment and children’s nutrition. It was not possible to compare microfinance models due to the small numbers of studies. More rigorous evidence is needed to evaluate the association between microfinance and social and health outcomes. PROSPERO registration number CRD42015026018.

• A critical evaluation of the limited evidence of the effects of microfinance on social and health outcomes. • Encompasses all regions of the low-and-middle income countries where microfinance is most likely to impact health and wellbeing of vulnerable populations. • Broad search terms used to capture all types of microfinance and a range of terminologies for the chosen outcomes. Rationale: Microfinance is the provision of financial services, including savings, deposit, and credit services, to the poor 1 . The term was first used in the early 1990s though schemes have been in operation in the developing world since the 1970s 2 . 'Microfinance' is subtly distinct from 'microcredit,' which refers to only small loans to poor people without a savings component.
Microfinance may also include provision of micro-insurance as an "add on" to the loans and saving component. Distinct characteristics of microfinance schemes are that they are short-term, have simple application procedures and do not require loan security but instead rely on a 'collective' guarantee from an enrolled group 3 . The purpose of microfinance is that the loans should reach the poor and move them out of poverty 4 .
The financial viability of microfinance programmes may be assessed by factors such as loan size, number of loans per person and repayment rates. One of the first studies to evaluate the economic impact of microfinance on participants was a quasi-experimental survey from Bangladesh 5 . This There is an emerging body of literature, including both experimental and quasi-experimental studies, looking at the social and health outcomes of microfinance programmes. In some cases, individual studies from the same region have reported contradictory results. For example, one study in Ghana demonstrated that combining microfinance and nutritional education led to improved indicators of childhood nutrition in the intervention group 7 , while a study in Ethiopia failed to demonstrate any difference in nutrition status between the children of clients and non-clients 8 . The two studies used different nutritional outcome measures as well as different age limits which makes synthesis of the findings difficult. Similarly, a study from Bangladesh reported improved female empowerment fifteen years later 9 , but there was no significant effect in a study in Hyderabad, India 10 . Most available studies are small and have insufficient power to detect small changes in outcomes.
Therefore, this systematic review brings together results from existing studies to assess whether Objectives: We aimed to evaluate the impact of microfinance schemes on social outcomes, specifically female contraceptive use and measures of female empowerment (intimate partner violence, decision making ability and mobility), as well as the effects on child nutrition.

METHODS
The protocol for this review is registered with PROSPERO, registration number CRD42015026018, and is available from http://www.crd.york.ac.uk/PROSPERO.
Eligibility Criteria: We included all controlled trials, observational studies, and analyses of panel data from South Asia, SSA and LAC 11 in women over the age of 15 and children under five. We included quasi-experimental studies (empirical studies used to estimate the causal impact of an intervention without randomisation). In most cases, panel data were longitudinal or "before and after" studies.
We also put in a geographical limitation to studies in countries within three World Bank regions with the highest number of developing countries 12 . Studies were included where the microfinance intervention comprised both savings and credit services, without physical collateral, to a poor or otherwise vulnerable population. Studies where microfinance was introduced and measured for expected change in outcome were included. Studies where an additional intervention was delivered in addition to microfinance were also included, provided that there was an intervention group where a microfinance intervention was assessed in comparison to the control group. In studies with more than one comparison group, the group without microfinance was considered as the main comparator. Studies were excluded where there were no suitable comparison data -either from a population who had not received microfinance, or pre-intervention data from those who went on to receive microfinance.
Outcome measures: Table 1 lists the outcome measures used to assess the impact of microfinance.
The Grameen foundation proposed three variables as indicators of the social performance of microfinance 13 : female use of contraceptives, female empowerment and childhood nutrition. Three indicators of female empowerment were chosen, self-reported intimate partner violence, decisionmaking ability and mobility. These were collated from indicators defined by the WHO 14 Millennium taskforce on gender equality 16 and also from literature on social measures of female empowerment [17][18][19][20][21] . The World Health Organisation (WHO) considers the health and wellbeing of women to be tied to their ability to access healthcare and have a say in decisions related to their health 22 . Improved health status could therefore be a possible consequence and proxy indicator of female empowerment. The WHO provides some standardised measures for use in assessing the health of women in a population. These include deaths from pregnancy-related complications, uptake of contraceptives and utilisation of perinatal services 22,23   Self-reported use of any contraceptive method to prevent or plan for pregnancy.

Female empowerment
Intimate Partner Violence (IPV): Self-reported intimate partner violence described as physical, sexual, or psychological harm by a current or former partner 25 . Mid-upper arm circumference (MUAC) -an absolute measure where a MUAC <11.5cm in children 6-60 months is considered as severe acute malnutrition (wasting) and 5cm moderate acute malnutrition 26 .
Information sources: EMBASE, MEDLINE, LILACS, CENTRAL and ECONLIT were searched from 1990 (when microfinance was first described 2 to 9 th September 2015. Theses were accessed through w w w . t h e s es . c om , and the references of included studies were tracked to identify other relevant papers. No language restrictions were applied. Searches were conducted using MESH headings and free text, as described in Supplement 1. Study selection, data extraction and quality assessment: Two authors (WG and LS) independently screened the titles and abstracts of retrieved studies against the study eligibility criteria.
Discrepancies were resolved by discussion and duplicates removed. Retrieved studies were translated into English, where necessary, and data were extracted by the two authors independently using a standard data extraction form. The methodological quality of included studies was assessed independently by WG and LS using the Cochrane Risk-of-Bias tool 27 for controlled trials and quasiexperimental studies and a modified Newcastle Ottawa Scale (NOS) 28 for cross-sectional surveys and analyses of panel data (Supplement 2).
Data synthesis and analysis: Meta-analyses were conducted using STATA 13 (StataCorp, College Station, TX) to pool the measures of effects from eligible studies. Where available, adjusted measures of effect were preferred over unadjusted measures. Statistical significance was set at a pvalue of <0·05. A random effects model was initially fitted for each meta-analysis. For studies with low heterogeneity analysis was repeated using a fixed effects model. Publication bias was assessed using funnel plots and Egger's asymmetry test (where at least five studies were available).
Descriptive synthesis was carried out where studies could not be meta-analysed.

RESULTS
Study selection: A total of 5416 titles were identified across the three groups of outcome measures, which reduced to 4821 after removal of duplicates. From these, 4584 titles were excluded as not being on microfinance as agreed mutually by two authors; 237 abstracts were subsequently screened. A total of 17 abstracts were translated for the authors to review. Each author screened the abstracts individually then came together to compare findings. The authors disagreed on two    Nature of the microfinance interventions evaluated: The most common microfinance model was group-lending as provided by formal microfinance institutions (MFIs) 9,10,[30][31][32][33][34]36,37,41,42,45,47 and community-based organisations (CBOs) 7,8,29,35,44 . MFIs required clients to be women above the age of eighteen, own less than 0·5 decimals of land (435 square feet) and have at least one household member in casual employment. Self-help groups and CBOs had fewer eligibility criteria but with greater emphasis on accumulation of savings 7,38,39,43,46 . In some studies microfinance was coupled with additional social and health interventions 7,29,32,35 .

Contraceptive Use
Three studies [29][30][31] evaluated the impact of microfinance on self-reported use of contraception using data from household cross-sectional surveys. One study 29 evaluated an intervention that combined microfinance with family planning education in Ethiopia. The other two 30,31 recruited clients from non-commercial MFIs in Bangladesh.
The impact of microfinance in the Ethiopian study was estimated at the level of the kebele (a cluster of villages) and showed no significant change in the proportion of married women reporting contraceptive use; individual-level estimates of the impact of microfinance were not available. A fixed-effects meta-analysis of individual-level data from the two Bangladeshi studies showed that women participating in microfinance were 53% more likely to report contraceptive use than nonclients [OR=1·53, 95%CI 1·21-1·94; Figure 2]. There was no heterogeneity between the studies, which is plausible given the similarity in the average age and socio-economic status of participants.

Intimate partner violence (IPV):
Four cross-sectional surveys [32][33][34]36 and one cluster RCT 35 reported this outcome. One survey 33 34 found that microfinance clients with secondary and higher education were 2-3 times more likely to experience IPV than comparable non-clients (p=<0·001), while wealthier clients were twice as likely to experience IPV than comparable nonclients (p=<0·001); there were no changes in exposure to IPV amongst the least educated and poorest groups.
A meta-analysis was not conducted due to high heterogeneity (I 2 =91·3%). This heterogeneity could have arisen because the threshold for reporting violence or the framing of the question may have differed between settings. The cluster RCT 35 was different both in design and in the add-on life skills training, which may have introduced further heterogeneity. The association between IPV and microfinance is therefore inconclusive.

Decision making agency:
Seven studies were included for this outcome, five from South Asia 37,39,40,42,45 and two from SSA 38,44 , with a similar number of cluster RCTs 37,38,44 and cross-sectional surveys 39,40,42,45 . This measure analysed a change from not being involved in decision making to being an active participant in household decisions. The outcome measures used were diverse and therefore unsuitable for meta-analysis. The results have been tabulated in more detail in Supplement 4 and include participation in financial and other household decisions (e.g. children's education and healthcare). Just over half the studies 37,40,42,45 showed a slightly higher degree of participation in certain household decisions by microfinance clients compared to non-clients. The other studies did not report any statistically significant changes. The impact of microfinance on women's decision making is therefore inconclusive.

Freedom to travel (mobility):
In the one study that assessed mobility, non-clients were more mobile than clients in one region, but in the two other regions studied the reverse was true 41 47 . Neither association was statistically significant. As the baseline group used was different and there were no raw data available, it was not possible to recalculate the ORs for pooling by meta-analysis.
One cross-sectional study found that the prevalence of malnutrition, based on HAZ-scores, was lower amongst children of microfinance clients than those of non-clients 46

DISCUSSION
Summary of evidence: Table 3 summarises the impact of microfinance across the three outcome domains based on the quantitative and qualitative syntheses described above.  Most studies showed no effect but a minority showed a significant positive effect on some areas of decision-making.

Mobility
No statistically significant impact.
Overall empowerment score Positive impact in two studies with mixed results and no change in two others.

Childhood nutrition
Positive impact in two of four studies, with no difference found in the remaining studies.
Fourteen of the 23 studies included in the review were from South Asia. This may limit the generalisability of the findings of this review to other geographical regions. However, this was expected as 84% of all microfinance clients are to be found in South Asia 48 .

Proposed mechanisms
Microfinance (whilst primarily improving economic stability) might empower women and improve child nutrition though a number of mechanisms. A small source of income, which is available primarily to the woman in the household, could increase the "bargaining power" of female

Contraceptive Use
Where individual-level data were available, the odds of reporting contraceptive use were higher in women participating in microfinance compared to those who did not. It has been argued that the women who self-select to join microfinance groups are more empowered than other women and this may in itself increase their likelihood of using contraception 4 . However, by comparing reported use in this group before and after the intervention 29,31 , it may be possible to demonstrate any effect attributable to microfinance, even with an inherent empowered state.

Other markers of female empowerment
Gender-related violence is known to be most commonly perpetuated by a person close to the woman, usually an intimate partner 49 . Female empowerment has been tied to the ability of a woman to report and speak up against such violence. A reduction in IPV is therefore one of the expected benefits of empowerment of women through microfinance. One cluster RCT 35 reported a reduction in IPV among microfinance clients. However, the combined microfinance with life skills training may have resulted in an intervention group different from the standard client therefore limiting the generalisability of their findings. The authors of this study argued that their training empowered the women to reveal IPV, therefore reducing underreporting 35 . Underreporting of IPV is common in many studies due to its sensitive nature 50 . Studies used trained local female interviewers to limit underreporting, but despite this, the response rate to IPV questions in one study was only 41% 34 .
Furthermore, women participating in microfinance may want to only highlight positive impacts of the intervention and not reveal any IPV. This raises ethical concerns that studies may fail to detect violence where it is actually present 50 , compromising efforts to highlight legitimate concerns of prevailing IPV.
In most cases, the decision-making ability of women participating in microfinance was not significantly different from that of non-clients. However, most studies analysed women's perceived decision-making ability, which may be different to their actual decision-making capability. In addition, composite indices of decision-making ability make it hard to untangle any impact of microfinance on decisions which are typically male-dominated (such as child marriage and education) and decisions which are traditionally less so (such as those related to the purchase of food). F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   18 Two studies 8,47 reported a lower likelihood of severe acute malnutrition in children of women participating in microfinance compared to non-participants, though were not statistically significant.

Childhood nutrition
Combining microfinance with nutritional education, as was the case in one study 7 , showed improvement in nutritional status in children of participating care-givers than non-participating caregivers. However, it is then difficult to isolate the specific effect of microfinance. In one SHG study 46 no attempt was made to adjust for other variables, such as household resources or education status, which may be a source of confounding.
Additionally, the inclusion of HAZ scores as a measure of nutritional status 46,47 in a cross-sectional study may be misleading. Height-for-age measures the effect of poor nutrition on the growth of a child. Growth faltering is slow in reversal and requires a longer follow-up period to detect 51 . It may be more prudent to use acute measures of malnutrition such as wasting (WHZ) which are likely to be more sensitive to change in nutritional status over shorter periods.

Strengths and limitations
Five comprehensive databases were searched in this review, including a large economic database.
The use of multiple indicators to measure women's empowerment and childhood nutrition also served to broaden the search to reduce the likelihood of missing relevant articles. The selection was carried out independently by two authors without any language restrictions, particularly important given the geographical regions studied.
The models used to deliver microfinance services varied across included studies. Some combined microfinance with education on family planning, 29 life skills 35 or health, nutrition and entrepreneurial skills, 7 which makes it difficult to evaluate the effect of microfinance alone. Although all interventions were taken to be similar for the purposes of this review, it is possible that the way the microfinance services were provided may have influenced the outcome. Given the small number of interventions of each type reviewed here, it is not possible to suggest a model of microfinance that is superior to others in terms of social performance.
In general, the most common source of bias in studies of the social impact of microfinance is selection bias, as participants self-select to either participate or not participate in the programme.
Whilst a cluster RCT might guard against selection bias, a recent study 10  included non-randomised studies in this review in order to not limit the evidence considered. The non-randomised studies included dealt with self-selection bias in two main ways, using either panel data in a quasi-experimental design or propensity score matching (PSM). However, additional analysis in of one of the studies included in this review suggested that the reduction in intimate partner violence demonstrated using conventional statistical methods did not hold when PSM was used 33 .
The average follow-up period of the studies included was three years. An alternative explanation for their statistically non-significant findings is that the observation period may have not been long enough to detect any change or may have missed any fleeting changes that occurred before the follow up survey. While changes in some measures of childhood malnutrition may be detectable within three years, changes in other outcomes requiring a shift in cultural and social norms may take much longer

AND
or "before and after" or "interrupted time series" or "time series" or "time-• OE] •-}OE -š]u ‰}]vš•_ Contraceptive Use microfinanc* or microcredit or microloan or "micro-financ*" or "micro-credit" or "micro-loan" or "micro financ*" or "micro credit" or "micro loan" or "small loan" or "small lend" or "micro vš OE‰OE]•Ž-}OE -u] OE} všOE ‰OE v µOE_ AND random* or "randomi*ed control trial" or "randomi*ed cluster trial" or study or analys* or cohort or "cross section*" or "cross-section*" or survey or "pre test and post test" or "pre-test and post-test" or "before and after" or "interrupted time series" or "time series" or "time-• OE] •-}OE -š]u ‰}]vš•_ AND "contraceptive*" or "contraception" or "reproductive" or gynaecolog* or gynecolog* or "birth control" or fertility Nutrition microfinanc* or microcredit or microloan or "micro-financ*" or "micro-credit" or "micro-loan" or "micro financ*" or "micro credit" or "micro loan" or "small loan" or "small lend" or "micro enterpris*" or -u] OE} všOE ‰OE v µOE_ AND nutrition OR malnutrition OR undernutrition OR under-nutrition OR underweight OR "MUAC" OR "mid-upper arm circumference" OR stunting OR "weight-for-age" OR "height for age" OR "heightfor-age" OR wasting OR whz OR "Z score" AND random* or "randomi*ed control trial" or "randomi*ed cluster trial" or study or analys* or cohort or "cross section*" or "cross-section*" or survey or "pre test and post test" or "pre-test and post-test" or "before and after" or "interrupted time series" or "time series" or "time-series" or "time points" IV CENTRAL microfinanc* or microcredit or microloan or "micro-financ*" or "micro-credit" or "micro-loan" or "micro financ*" or "micro credit" or "micro loan" or "small loan" or "small lend" or "micro enterpris*" or "micro enterpreneur"  1) The subjects in different outcome groups are comparable, based on the study design or analysis.
Confounding factors are controlled.
a) The study controls for the most important factors t age, education level, social status (select one). * * b) The study displays data on the above factors comparing intervention and nonintervention groups but does not adjust*  The statistical test used to analyze the data is clearly described and appropriate, and the measurement of the association is presented, including confidence intervals and the probability level The statistical test is not appropriate, not described or incomplete.

Total # of stars: /11
This scale has been adapted from the Newcastle-Ottawa Quality Assessment Scale for cohort studies to perform a quality assessment of cross-sectional studies for this systematic review.

5
Eligibility criteria 6 Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered, language, publication status) used as criteria for eligibility, giving rationale.

5
Information sources 7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and date last searched. 6 Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated.

Supplement 1
Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable, included in the meta-analysis).

7
Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators.

7
Data items 11 List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made. 14 Additional analyses 16 Describe methods of additional analyses (e.g., sensitivity or subgroup analyses, meta-regression), if done, indicating which were pre-specified.

Study selection
17 Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally with a flow diagram. Figure  1]

[
Study characteristics 18 For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and provide the citations. Table 2 Risk of bias within studies 19 Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12).

-11
[ Table 2] Results of individual studies 20 For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention group (b) effect estimates and confidence intervals, ideally with a forest plot.

-14
Synthesis of results 21 Present results of each meta-analysis done, including confidence intervals and measures of consistency.

DISCUSSION
Summary of evidence 24 Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., healthcare providers, users, and policy makers).

15
Limitations 25 Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified research, reporting bias).

Data sharing agreement
This is a secondary analysis of published data. We do not hold any unpublished data from the study. Further information about the data analysis can be obtained by contacting the corresponding author.

Results:
We included 27 studies. Microfinance was associated with a 64% increase in the number of women using contraceptives [OR 1·64, 95%CI 1.45 1.86]. We found mixed results for the association between microfinance and intimate partner violence. Some positive changes were noted in female empowerment. Improvements in children's nutrition were noted in three studies.

Conclusion:
Microfinance has the potential to generate changes in contraceptive use, female empowerment and children's nutrition. It was not possible to compare microfinance models due to the small numbers of studies. More rigorous evidence is needed to evaluate the association between microfinance and social and health outcomes.
Funding: WG received a Commonwealth Scholarship to support her study. There was no other funding for this work.

Strengths of the study:
A critical evaluation of the limited evidence of the effects of microfinance on social and health outcomes.
Encompasses all regions of the low-and-middle income countries where microfinance is most likely to impact health and wellbeing of vulnerable populations.
Broad search terms used to capture all types of microfinance and a range of terminologies for the chosen outcomes.
No language restrictions -captured all Latin American literature which is vital in the field of microfinance.

Limitations of the study:
We found few randomised controlled trials in the field and relied upon the inclusion of quasi-experimental studies. simple application procedures and do not require loan security but instead rely on a 'collective' guarantee from an enrolled group 3 . The purpose of microfinance is that the loans should reach the poor and move them out of The two studies used different nutritional outcome measures as well as different age limits which makes synthesis of the findings difficult. Similarly, a study from Bangladesh reported improved female empowerment fifteen years later 9 , but there was no significant effect in a study in Hyderabad, India 10 . Most available studies are small and have insufficient power to detect small changes in outcomes. Therefore, this systematic review brings together results from existing studies to assess whether receiving microfinance is associated with changes in women's empowerment and the well-being of their children.
Objectives: We aimed to evaluate the impact of microfinance schemes on health and social outcomes, specifically female contraceptive use and measures of female empowerment (intimate partner violence, decision making ability and mobility), as well as the effects on child nutrition. Asia, SSA and LAC 11 in women over the age of 15 and children under five. We included quasi-experimental studies (empirical studies used to estimate the causal impact of an intervention without randomisation). In most cases, panel data were longitudinal or "before and after" studies. We also put in a geographical limitation to studies in countries within three World Bank regions with the highest number of developing countries 12 . Studies were included where the microfinance intervention comprised both savings and credit services, without physical collateral, to a poor or otherwise vulnerable population. Studies where microfinance was introduced and measured for expected change in outcome were included. Studies where an additional intervention was delivered in addition to microfinance were also included, provided that there was an intervention group where a microfinance intervention was assessed in comparison to the control group. In studies with more than one comparison group, the group without microfinance was considered as the main comparator. Studies were excluded where there were no suitable comparison data -either from a population who had not received microfinance, or pre-intervention data from those who went on to receive microfinance.
Patient and Public Involvement: There was no PPI involvement in the design or conduct of this review. The results were presented and discussed at a dissemination workshop in Patna, Bihar.
We conducted a workshop "Women's Empowerment and Child Health: Exploring the Impact of Rojiroti Microfinance in Poor Communities in Bihar-An Indo-UK collaboration" in Patna, India on May 22, 2018. It was attended by more than 30 women who participate in microfinance, and a wide range of local stakeholders. The results of this review and other work were presented and discussed it this meeting and women's views were noted to enable further research in this area.
Outcome measures: Table 1 lists the outcome measures used to assess the impact of microfinance. The Grameen foundation proposed three variables as indicators of the social performance of microfinance 13 : female use of contraceptives, female empowerment and children's nutrition. [14][15][16][17][18][19] The World Health Organisation (WHO) considers the health and wellbeing of women to be tied to their ability to access healthcare and have a say in decisions related to their health 14 . Improved health status could therefore be a possible consequence and proxy indicator of female empowerment. The WHO provides some standardised measures for use in assessing the health of women in a population. These include deaths from pregnancyrelated complications, uptake of contraceptives and utilisation of perinatal services 14 15 . Uptake of contraceptives is one of the measures proposed by the Grameen Foundation. 16 Due to the broadness of the term "female empowerment", indicators collated from definitions used by the WHO 14 15 and the UN Millennium taskforce on gender equality 16 and also from literature on social measures of female empowerment 17 19 were used to inform the selection of the three outcome measures of female empowerment used in this systematic review. These were self-reported intimate partner violence, decision-making ability and mobility. Self-reported use of any contraceptive method to prevent or plan for pregnancy.

Female empowerment
Intimate Partner Violence (IPV): Self-reported intimate partner violence described as physical, sexual, or psychological harm by a current or former partner 20 .
Sole decision-making ability: Self-reported independent decision-making ability where the woman is not the head of household; including but not limited to, household expenditure, children's education or as a combined measure of empowerment as defined by individual study authors.
Mobility: Self-reported freedom to travel out of the village or to attend social events without the permission or accompaniment of a male relative.

Children's nutrition
Standard nutritional measures for children aged <5 as defined by the WHO Global Database on Child Growth and Malnutrition (WHO). Moderate undernutrition (malnutrition) was defined as a Zscore <-2 but > -3 standard deviations (SD) from the mean. Severe undernutrition (malnutrition) was defined as a Z-score<-3 SD from the mean.

Weight-for-age Z-score (WAZ)
Height (or length)-for-age Z-score (HAZ) -the most indicative measure of chronic undernutrition over a prolonged period leading to growth retardation known as stunting.
Weight-for-height (or length) (WHZ) -most indicative measure of acute undernutrition known as wasting. This distinguishes short children of normal weight and tall children of low weight that may not be captured by WAZ or HAZ.
Information sources: EMBASE, MEDLINE, LILACS, CENTRAL and ECONLIT were searched from 1990 (when microfinance was first described 2 to 9 th September 2015. These were accessed through w w w . t h es es . c om , and the references of included studies were tracked to identify other relevant papers. No language restrictions were applied. Searches were conducted using MESH headings and free text, as described in Supplement 1.  Publication bias was assessed using funnel plots and Egger's asymmetry test (where at least five studies were available). Descriptive synthesis was carried out where studies could not be meta-analysed.

RESULTS
Study selection: A total of 5659 titles were identified across the three groups of outcome measures, which reduced to 5298 after removal of duplicates. From these, 5023 titles were excluded as not being on microfinance as agreed mutually by two authors; 275 abstracts were subsequently screened. A total of 17 abstracts were translated for the authors to review. Each author screened the abstracts individually then came together to compare findings. The  Table 2 summarises the characteristics of the included studies.

Contraceptive Use
Four studies 5 23 25 26 evaluated the impact of microfinance on self-reported use of contraception using data from household cross-sectional surveys. One study 23 evaluated an intervention that combined microfinance with family planning education in Ethiopia. The other 3 studies 24-26 recruited clients from non-commercial MFIs in Bangladesh.
The impact of microfinance in the Ethiopian study was estimated at the level of the kebele (a cluster of villages) and showed no significant change in the proportion of married women reporting contraceptive use; individual-level estimates of the impact of microfinance were not available. A fixed-effects meta-analysis of individual-level data from the three Bangladeshi studies showed that women participating in microfinance were 64% more likely to report contraceptive use than nonclients [OR=1.64, 95%CI 1.45 1.86; Figure 2]. There was no heterogeneity between the studies, which is plausible given the similarity in the average age and socio-economic status of participants.

Freedom to travel (mobility):
In the one study that assessed mobility, non-clients were more mobile than clients in one region, but in the two other regions studied the reverse was true 38 . No formal statistical comparisons between groups were presented.

Children's nutrition
Five studies, 4 from SSA 7 8 43 44 and 1 from India 45 , evaluated the effect of microfinance on children's nutrition. Three 8 43 44 were cross-sectional surveys, 1 was a quasi-experimental study with a 16 month follow-up period 7 while 1 was a cluster randomised controlled trial 45 . Two studies 7 44 included only children between 6-36 months of age while the other 3 included children under five years.  44 . Neither association was statistically significant. As the baseline group used was different and there were no raw data available, it was not possible to recalculate the ORs for pooling by meta-analysis.
One cross-sectional study found that the prevalence of malnutrition, based on HAZ-scores, was lower amongst children of microfinance clients than those of non-clients 43

DISCUSSION
Summary of evidence: Table 3 summarises the impact of microfinance across the three outcome domains based on the quantitative and qualitative syntheses described above. Decision making ability Most studies showed no effect but a minority showed a significant positive effect on some areas of decision-making.

Mobility
No statistically significant impact.
Overall empowerment score Positive impact in two studies with mixed results and no change in two others.
Children's nutrition Positive impact in three of five studies, with no difference found in the remaining studies.
Seventeen of the 27 studies included in the review were from South Asia. This may limit the generalisability of the findings of this review to other geographical regions. However, this was expected as 84% of all microfinance clients are to be found in South Asia 46 .

Proposed mechanisms
Microfinance (whilst primarily improving economic stability) might empower women and improve child nutrition though a number of mechanisms. A small source of income, which is available primarily to the woman in the household, could increase the "bargaining power" of female participants, in household decision making. Peer support and shared learning from other participants might have a similar effect. We have chosen the outcome measures most likely to reflect this increased bargaining power, including a woman's decisions about contraception and her self-reported empowerment. Furthermore, that women are often the primary household decisionmakers on issues such as buying food (which will affect child nutrition) and on access to healthcare for children. These factors could interact to enable women to overcome social, cultural and economic barriers that affect their status (Figure 3)

Contraceptive Use
Where individual-level data were available, the odds of reporting contraceptive use were higher in women participating in microfinance compared to those who did not. It has been argued that the women who self-select to join microfinance groups are more empowered than other women and this may in itself increase their likelihood of using contraception 4 . However, by comparing reported use in this group before and after the intervention 23 25 , it is possible to demonstrate a positive effect attributable to microfinance, even with an inherent empowered state.

Intimate Partner Violence
Gender-related violence is known to be most commonly perpetuated by a person close to the woman, usually an intimate partner 46 . Although a reduction in IPV is one of the expected benefits of empowerment of women through microfinance, empowerment may also enable women to report more IPV, thus increasing the rate of reported IPV. One cluster RCT 31 reported a reduction in IPV among microfinance clients. However, the combined microfinance with life skills training may have resulted in an intervention group different from the standard client therefore limiting the generalisability of their findings. The authors of this study argued that their training empowered the women to reveal IPV, therefore reducing underreporting 31 . Underreporting of IPV is common in many studies due to its sensitive nature 47 . Studies used trained local female interviewers to limit underreporting, but despite this, the response rate to IPV questions in one study was only 41% 29 .
Furthermore, women participating in microfinance may want to only highlight positive impacts of the intervention and not reveal any IPV. This raises ethical concerns that studies may fail to detect violence where it is actually present 47 .
Studies that have reported increase in IPV linked to microfinance programmes 29 have also argued that microfinance loans may have caused more economic stress in the family leading to greater occasions for conflict. Some authors explain this as the "status inconsistency theory" where in status differentials may lead to dysfunctional behaviour when and individual who expects to have a higher status in a relationship is threatened by the increase in the status of another 30 . Previously there may have been fewer conflicts as the man would have managed finances single-handedly while with empowerment, the wife becomes involved in these decisions, generating more occasions where conflict leading to IPV could occur.

Decision Making Ability
In most cases, the decision-making ability of women participating in microfinance was not significantly different from that of non-clients. However, most studies analysed women's perceived decision-making ability, which may be different to their actual decision-making capability. In addition, composite indices of decision-making ability make it hard to untangle any impact of

Children's nutrition
Three studies 8 44 45 reported a lower likelihood of severe acute malnutrition in children of women participating in microfinance compared to non-participants, including one that showed a statistically significant reduction in malnutrition 45 . Combining microfinance with nutritional education, as was the case in one study 7 , showed improvement in nutritional status in children of participating caregivers than non-participating care-givers. However, it is then difficult to isolate the specific effect of microfinance. In one SHG study 43 no attempt was made to adjust for other variables, such as household resources or education status, which may be a source of confounding.
Additionally, the inclusion of HAZ scores as a measure of nutritional status 43

Strengths and limitations
Five comprehensive databases were searched in this review, including a large economic database.
The use of multiple indicators to measure women's empowerment and children's nutrition also served to broaden the search to reduce the likelihood of missing relevant articles. The selection was carried out independently by two authors without any language restrictions, particularly important given the geographical regions studied.
The models used to deliver microfinance services varied across included studies. Some combined microfinance with education on family planning 23  In general, the most common source of bias in studies of the social impact of microfinance is selection bias, as participants self-select to either participate or not participate in the programme.
Although, it may be argued that it would be difficult to randomise people to microfinance as the intervention may not be desired by all; therefore measuring effectiveness in those who did not desire it to begin with, may be problematic. Whilst a cluster RCT might guard against selection bias, a recent study 10 highlighted the current challenge in achieving randomisation due to the widespread diffusion of microfinance in some regions of South Asia leading to difficulties in identifying unexposed control clusters. Therefore, we included non-randomised studies in this review in order to not limit the evidence considered. The non-randomised studies included dealt with self-selection bias in two main ways, using either panel data in a quasi-experimental design or propensity score matching (PSM). However, additional analysis in of one of the studies included in this review suggested that the reduction in intimate partner violence demonstrated using conventional statistical methods did not hold when PSM was used 28 .
The average follow-up period of the studies included was three years. An alternative explanation for their statistically non-significant findings is that the observation period may have not been long enough to detect any change or may have missed any fleeting changes that occurred before the follow up survey. While changes in some measures of children's malnutrition may be detectable within three years, changes in other outcomes requiring a shift in cultural and social norms may take much longer

Conclusions
In conclusion, our findings suggest that for the types of microfinance interventions assessed in this study, there may be an association between microfinance and increasing contraceptive use,     (microfinanc* or microcredit or microloan or "micro-financ*" or "micro-credit" or "micro-loan" or "micro financ*" or "micro credit" or "micro loan" or "small loan" or "small lend" or "micro enterpris*" or "micro enterpreneur" 2. (random* or "randomi*ed control trial" or "randomi*ed cluster trial" or study or analys* or cohort or "cross section*" or "cross-section*" or survey or " pre test and post test" or "pre-test and posttest" or "before and after" or "interrupted time series" or "time series" or "time-series" or "time points" (microfinanc* or microcredit or microloan or "micro-financ*" or "micro-credit" or "micro-loan" or "micro financ*" or "micro credit" or "micro loan" or "small loan" or "small lend" or "micro enterpris*" or "micro enterpreneur").mp. [mp=title, abstract, original title, name of substance word, 3. ("health" or "outcome" or "evaluat*" or "intervention" or "impact" or "result*" or "effect*").mp.

Total # of stars: /11
This scale has been adapted from the Newcastle-Ottawa Quality Assessment Scale for cohort studies to perform a quality assessment of cross-sectional studies for this systematic review. Women's empowerment as measured by intimate partner violence, decision making agency, mobility and Women's empowerment as measured by intimate partner violence, decision making agency, mobility and distinctly as uptake of a contraceptive method. Empowerment of women may also be linked to improved distinctly as uptake of a contraceptive method. Empowerment of women may also be linked to improved childhood nutrition. This will be measured by weight-for-age Z-scores, height-for-age Z-scores, weight-for-childhood nutrition. This will be measured by weight-for-age Z-scores, height-for-age Z-scores, weight-forheight Z scores and mid-upper arm circumference. height Z scores and mid-upper arm circumference.

Participants/population Participants/population
Inclusion: Women above the age of fifteen and children under-five for the outcome on childhood nutrition. Inclusion: Women above the age of fifteen and children under-five for the outcome on childhood nutrition.
Exclusion: Men, children above five years Exclusion: Men, children above five years

Intervention(s), exposure(s) Intervention(s), exposure(s)
Intervention: Microfinance schemes defined as a combination of savings and credit services offered without Intervention: Microfinance schemes defined as a combination of savings and credit services offered without physical collateral to a population thought to be poor or otherwise vulnerable through any organisation or physical collateral to a population thought to be poor or otherwise vulnerable through any organisation or institution. institution.
The provider may be non-profit, e.g. NGO, self-help group (SHG), community-based organisation or The provider may be non-profit, e.g. NGO, self-help group (SHG), community-based organisation or microfinance bank, or a for-profit micro-finance institution, e.g., commercial bank. microfinance bank, or a for-profit micro-finance institution, e.g., commercial bank.
Studies having an additional intervention will also be considered, provided that the primary intervention is Studies having an additional intervention will also be considered, provided that the primary intervention is microfinance. microfinance.

Comparator(s)/control Comparator(s)/control
Populations without any microfinance services or the same population prior to receiving microfinance. In studies Populations without any microfinance services or the same population prior to receiving microfinance. In studies with more than one comparator group, the group without microfinance will be considered as the main with more than one comparator group, the group without microfinance will be considered as the main comparator. comparator.

Context Context
Developing

Secondary outcome(s) Secondary outcome(s)
None None

Data extraction (selection and coding) Data extraction (selection and coding)
The search will be conducted and subsequent papers reviewed for eligibility independently by two researchers The search will be conducted and subsequent papers reviewed for eligibility independently by two researchers in three stages; title, abstract and full-text. in three stages; title, abstract and full-text. A data extraction form will be completed for each selected study by one researcher under the following sub-A data extraction form will be completed for each selected study by one researcher under the following subheadings; publication details, study details, nature of study, intervention and results. The data extraction forms headings; publication details, study details, nature of study, intervention and results. The data extraction forms will then be reviewed by the second researcher. will then be reviewed by the second researcher. This is to be used in further analysis and synthesis of the data. This is to be used in further analysis and synthesis of the data.
Any disparities will be solved be mutual consensus between the two primary researchers. If this is not possible, Any disparities will be solved be mutual consensus between the two primary researchers. If this is not possible, the input of the third researcher will be sought. the input of the third researcher will be sought.

Risk of bias (quality) assessment Risk of bias (quality) assessment
The selected studies will be assessed for risk of bias by two researchers using the Cochrane Collaboration's The selected studies will be assessed for risk of bias by two researchers using the Cochrane Collaboration's tool for assessing risk of bias in randomised controlled trials and for quality by the Newcastle-Ottawa Quality tool for assessing risk of bias in randomised controlled trials and for quality by the Newcastle-Ottawa Quality Assessment Scale in non-randomised studies. Assessment Scale in non-randomised studies.
Any disparities will be resolved be mutual consensus between the two primary researchers. If this is not Any disparities will be resolved be mutual consensus between the two primary researchers. If this is not possible, the input of the third researcher will be sought. possible, the input of the third researcher will be sought.

Strategy for data synthesis Strategy for data synthesis
Outcome measures will be extracted from the studies and used in the meta-analyses. The studies providing an Outcome measures will be extracted from the studies and used in the meta-analyses. The studies providing an appropriate measure of effect will be weighted using a quality rating system and then stratified by quality score. appropriate measure of effect will be weighted using a quality rating system and then stratified by quality score.
A descriptive analysis will be done for studies providing quantitative outcome measures not suitable for meta-A descriptive analysis will be done for studies providing quantitative outcome measures not suitable for metaanalysis. analysis.
A fixed-effects or a random-effects model will be used in pooling of the data and a suitable method of estimating A fixed-effects or a random-effects model will be used in pooling of the data and a suitable method of estimating variance in studies will be applied. variance in studies will be applied. The summary estimate of the effect size will be done in each stratum The summary estimate of the effect size will be done in each stratum according to quality score, i.e. high, medium and low quality score, and statistical tests (I-squared) used to according to quality score, i.e. high, medium and low quality score, and statistical tests (I-squared) used to check for heterogeneity. check for heterogeneity.

Analysis of subgroups or subsets Analysis of subgroups or subsets
A sub-group analysis of the measures of effect chosen will be done according to region to detect any variations A sub-group analysis of the measures of effect chosen will be done according to region to detect any variations between regions. The three regions will be geographically specified as Sub-Saharan Africa, South Asia and between regions. The three regions will be geographically specified as Sub-Saharan Africa, South Asia and South America. The results will be presented by tables within the text of the review or if possible in forest plots South America. The results will be presented by tables within the text of the review or if possible in forest plots in the meta-analysis in the meta-analysis

Conflicts of interest Conflicts of interest
None known None known

Language Language
English English

Country Country
England England

Revision note for this version Revision note for this version
Update to reflect the completion of the review. Update to reflect the completion of the review. PROSPERO PROSPERO This information has been provided by the named contact for this review. CRD has accepted this information in good faith This information has been provided by the named contact for this review. CRD has accepted this information in good faith and registered the review in PROSPERO. CRD bears no responsibility or liability for the content of this registration and registered the review in PROSPERO. CRD bears no responsibility or liability for the content of this registration record, any associated files or external websites. record, any associated files or external websites. 14 Additional analyses 16 Describe methods of additional analyses (e.g., sensitivity or subgroup analyses, meta-regression), if done, indicating which were pre-specified.

Study selection
17 Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally with a flow diagram. Figure  1]

[
Study characteristics 18 For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and provide the citations. Table 2 Risk of bias within studies 19 Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12).

-11 [Table 2]
Results of individual studies 20 For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention group (b) effect estimates and confidence intervals, ideally with a forest plot.

-14
Synthesis of results 21 Present results of each meta-analysis done, including confidence intervals and measures of consistency.

DISCUSSION
Summary of evidence 24 Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., healthcare providers, users, and policy makers).

15
Limitations 25 Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified research, reporting bias).

Competing interests statement
There are no competing interests for any authors.

Data sharing agreement
This is a secondary analysis of published data. We do not hold any unpublished data from the study. Further information about the data analysis can be obtained by contacting the corresponding author.

Results:
We included 27 studies. Microfinance was associated with a 64% increase in the number of women using contraceptives [OR 1·64, 95%CI 1.45 1.86]. We found mixed results for the association between microfinance and intimate partner violence. Some positive changes were noted in female empowerment. Improvements in children's nutrition were noted in three studies.

Strengths of the study:
A critical evaluation of the limited evidence of the effects of microfinance on social and health outcomes.
Encompasses all regions of the low-and-middle income countries where microfinance is most likely to impact health and wellbeing of vulnerable populations.
Broad search terms used to capture all types of microfinance and a range of terminologies for the chosen outcomes.
No language restrictions -captured all Latin American literature which is vital in the field of microfinance.

Limitations of the study:
We found few randomised controlled trials in the field and relied upon the inclusion of quasi-experimental studies.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  The two studies used different nutritional outcome measures as well as different age limits which makes synthesis of the findings difficult. Similarly, a study from Bangladesh reported improved female empowerment fifteen years later 9 , but there was no significant effect in a study in Hyderabad, India 10 . Most available studies are small and have insufficient power to detect small changes in outcomes. Therefore, this systematic review brings together results from existing studies to assess whether receiving microfinance is associated with changes in women's empowerment and the well-being of their children.
Objectives: We aimed to evaluate the impact of microfinance schemes on health and social outcomes, specifically female contraceptive use and measures of female empowerment (intimate partner violence, decision making ability and mobility), as well as the effects on child nutrition.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  Asia, SSA and LAC 11 in women over the age of 15 and children under five. We included quasi-experimental studies (empirical studies used to estimate the causal impact of an intervention without randomisation). In most cases, panel data were longitudinal or "before and after" studies. We also put in a geographical limitation to studies in countries within three World Bank regions with the highest number of developing countries 12 . Studies were included where the microfinance intervention comprised both savings and credit services, without physical collateral, to a poor or otherwise vulnerable population. Studies where microfinance was introduced and measured for expected change in outcome were included. Studies where an additional intervention was delivered in addition to microfinance were also included, provided that there was an intervention group where a microfinance intervention was assessed in comparison to the control group. In studies with more than one comparison group, the group without microfinance was considered as the main comparator. Studies were excluded where there were no suitable comparison data -either from a population who had not received microfinance, or pre-intervention data from those who went on to receive microfinance.
Patient and Public Involvement: There was no PPI involvement in the design or conduct of this review. The results were presented and discussed at a dissemination workshop in Patna, Bihar.
We conducted a workshop "Women's Empowerment and Child Health: Exploring the Impact of Rojiroti Microfinance in Poor Communities in Bihar-An Indo-UK collaboration" in Patna, India on May 22, 2018. It was attended by more than 30 women who participate in microfinance, and a wide range of local stakeholders. The results of this review and other work were presented and discussed it this meeting and women's views were noted to enable further research in this area.
Outcome measures: Table 1 lists the outcome measures used to assess the impact of microfinance. The Grameen foundation proposed three variables as indicators of the social performance of microfinance 13 : female use of contraceptives, female empowerment and children's nutrition. [14][15][16][17][18][19] The World Health Organisation (WHO) considers the health and wellbeing of women to be tied to their ability to access healthcare and have a say in decisions related to their health 14 . Improved health status could therefore be a possible consequence and proxy indicator of female empowerment. The WHO provides some standardised measures for use in assessing the health of women in a population. These include deaths from pregnancyrelated complications, uptake of contraceptives and utilisation of perinatal services 14 15 . Uptake of contraceptives is one of the measures proposed by the Grameen Foundation. 16 Due to the broadness of the term "female empowerment", indicators collated from definitions used by the WHO 14 15 and the UN Millennium taskforce on gender equality 16 and also from literature on social measures of female empowerment 17 19 were used to inform the selection of the three outcome measures of female empowerment used in this systematic review. These were self-reported intimate partner violence, decision-making ability and mobility. Self-reported use of any contraceptive method to prevent or plan for pregnancy.

Female empowerment
Intimate Partner Violence (IPV): Self-reported intimate partner violence described as physical, sexual, or psychological harm by a current or former partner 20 .
Sole decision-making ability: Self-reported independent decision-making ability where the woman is not the head of household; including but not limited to, household expenditure, children's education or as a combined measure of empowerment as defined by individual study authors.
Mobility: Self-reported freedom to travel out of the village or to attend social events without the permission or accompaniment of a male relative.

Children's nutrition
Standard nutritional measures for children aged <5 as defined by the WHO Global Database on Child Growth and Malnutrition (WHO). Moderate undernutrition (malnutrition) was defined as a Zscore <-2 but > -3 standard deviations (SD) from the mean. Severe undernutrition (malnutrition) was defined as a Z-score<-3 SD from the mean.

Weight-for-age Z-score (WAZ)
Height (or length)-for-age Z-score (HAZ) -the most indicative measure of chronic undernutrition over a prolonged period leading to growth retardation known as stunting.
Weight-for-height (or length) (WHZ) -most indicative measure of acute undernutrition known as wasting. This distinguishes short children of normal weight and tall children of low weight that may not be captured by WAZ or HAZ.

Contraceptive Use
Four studies 5 23 25 26 evaluated the impact of microfinance on self-reported use of contraception using data from household cross-sectional surveys. One study 23  women participating in microfinance were 64% more likely to report contraceptive use than nonclients [OR=1.64, 95%CI 1.45 1.86; Figure 2]. There was no heterogeneity between the studies, which is plausible given the similarity in the average age and socio-economic status of participants.

Female empowerment
Seventeen studies evaluated the impact of microfinance on female empowerment. Eight were conventional cross-sectional studies 27-30 32 36 37 39 42 , 3 were quasi-experimental 9 38 40 and 6 were cluster-randomised controlled trials (cluster RCTs) 10 29 found that microfinance clients with secondary and higher education were 2-3 times more likely to experience IPV than comparable non-clients (p=<0·001), while wealthier clients were twice as likely to experience IPV than comparable non-clients (p=<0·001); there were no changes in exposure to IPV amongst the least educated and poorest groups. This finding was confirmed by Murshid et al. 30

Freedom to travel (mobility):
In the one study that assessed mobility, non-clients were more mobile than clients in one region, but in the two other regions studied the reverse was true 38 . No formal statistical comparisons between groups were presented.

Children's nutrition
Five studies, 4 from SSA 7 8 43 44 and 1 from India 45 , evaluated the effect of microfinance on children's nutrition. Three 8 43 44 were cross-sectional surveys, 1 was a quasi-experimental study with a 16 month follow-up period 7 while 1 was a cluster randomised controlled trial 45 . Two studies 7 44 included only children between 6-36 months of age while the other 3 included children under five years.  44 . Neither association was statistically significant. As the baseline group used was different and there were no raw data available, it was not possible to recalculate the ORs for pooling by meta-analysis.
One cross-sectional study found that the prevalence of malnutrition, based on HAZ-scores, was lower amongst children of microfinance clients than those of non-clients 43

DISCUSSION
Summary of evidence: Table 3 summarises the impact of microfinance across the three outcome domains based on the quantitative and qualitative syntheses described above. Decision making ability Most studies showed no effect but a minority showed a significant positive effect on some areas of decision-making.

Mobility
No statistically significant impact.
Overall empowerment score Positive impact in two studies with mixed results and no change in two others.
Children's nutrition Positive impact in three of five studies, with no difference found in the remaining studies.
Seventeen of the 27 studies included in the review were from South Asia. This may limit the generalisability of the findings of this review to other geographical regions. However, this was expected as 84% of all microfinance clients are to be found in South Asia 46 . Other included studies, nine from Africa and one from Latin America, are geographically heterogeneous but catered to women of a similar economic background. These populations are potentially comparable for the purposes of a study looking at the impact of microfinance. However, it is of note that the review includes populations from a wider geographical range, with diverse political, cultural and social backgrounds.

Proposed mechanisms
Microfinance (whilst primarily improving economic stability) might empower women and improve child nutrition though a number of mechanisms. A small source of income, which is available primarily to the woman in the household, could increase the "bargaining power" of female participants, in household decision making. Peer support and shared learning from other participants might have a similar effect. We have chosen the outcome measures most likely to reflect this increased bargaining power, including a woman's decisions about contraception and her self-reported empowerment. Furthermore, that women are often the primary household decisionmakers on issues such as buying food (which will affect child nutrition) and on access to healthcare for children. These factors could interact to enable women to overcome social, cultural and economic barriers that affect their status (Figure 3)

Contraceptive Use
Where individual-level data were available, the odds of reporting contraceptive use were higher in women participating in microfinance compared to those who did not. It has been argued that the women who self-select to join microfinance groups are more empowered than other women and this may in itself increase their likelihood of using contraception 4 . However, by comparing reported use in this group before and after the intervention 23 25 , it is possible to demonstrate a positive effect attributable to microfinance, even with an inherent empowered state.

Intimate Partner Violence
Gender-related violence is known to be most commonly perpetuated by a person close to the woman, usually an intimate partner 46 . Although a reduction in IPV is one of the expected benefits of empowerment of women through microfinance, empowerment may also enable women to report more IPV, thus increasing the rate of reported IPV. One cluster RCT 31 reported a reduction in IPV among microfinance clients. However, the combined microfinance with life skills training may have resulted in an intervention group different from the standard client therefore limiting the generalisability of their findings. The authors of this study argued that their training empowered the women to reveal IPV, therefore reducing underreporting 31 . Underreporting of IPV is common in many studies due to its sensitive nature 47 . Studies used trained local female interviewers to limit underreporting, but despite this, the response rate to IPV questions in one study was only 41% 29 .
Furthermore, women participating in microfinance may want to only highlight positive impacts of the intervention and not reveal any IPV. This raises ethical concerns that studies may fail to detect violence where it is actually present 47 .
Studies that have reported increase in IPV linked to microfinance programmes 29 have also argued that microfinance loans may have caused more economic stress in the family leading to greater occasions for conflict. Some authors explain this as the "status inconsistency theory" where in status differentials may lead to dysfunctional behaviour when and individual who expects to have a higher status in a relationship is threatened by the increase in the status of another 30 . Previously there may have been fewer conflicts as the man would have managed finances single-handedly while with empowerment, the wife becomes involved in these decisions, generating more occasions where conflict leading to IPV could occur.

Decision Making Ability
In most cases, the decision-making ability of women participating in microfinance was not significantly different from that of non-clients. However, most studies analysed women's perceived decision-making ability, which may be different to their actual decision-making capability. In addition, composite indices of decision-making ability make it hard to untangle any impact of microfinance on decisions which are typically male-dominated (such as child marriage and education) and decisions which are traditionally less so (such as those related to the purchase of food).

Children's nutrition
Three studies 8 44 45 reported a lower likelihood of severe acute malnutrition in children of women participating in microfinance compared to non-participants, including one that showed a statistically significant reduction in malnutrition 45 . Combining microfinance with nutritional education, as was the case in one study 7 , showed improvement in nutritional status in children of participating caregivers than non-participating care-givers. However, it is then difficult to isolate the specific effect of microfinance. In one SHG study 43 no attempt was made to adjust for other variables, such as household resources or education status, which may be a source of confounding.
Additionally, the inclusion of HAZ scores as a measure of nutritional status 43 44 in a cross-sectional study may be misleading. In their cluster randomised trial, Ojha et al. report an improvement in all other indices of malnutrition other than HAZ and stunting after an 18 month period 45 . Height-for-age measures the effect of poor nutrition on the growth of a child. Growth faltering is slow in reversal and requires a longer follow-up period to detect 48 . It may be more prudent to use acute measures of malnutrition such as wasting (WHZ) which are likely to be more sensitive to change in nutritional status over shorter periods.

Strengths and limitations
Five comprehensive databases were searched in this review, including a large economic database.
The use of multiple indicators to measure women's empowerment and children's nutrition also served to broaden the search to reduce the likelihood of missing relevant articles. The selection was carried out independently by two authors without any language restrictions, particularly important given the geographical regions studied.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   20 The models used to deliver microfinance services varied across included studies. Some combined microfinance with education on family planning 23 , life skills 31 or health, nutrition and entrepreneurial skills 7 , which makes it difficult to evaluate the effect of microfinance alone. Although all interventions were taken to be similar for the purposes of this review, it is possible that the way the microfinance services were provided may have influenced the outcome. Given the small number of interventions of each type reviewed here, it is not possible to suggest a model of microfinance that is superior to others in terms of social performance.
In general, the most common source of bias in studies of the social impact of microfinance is selection bias, as participants self-select to either participate or not participate in the programme.
Although, it may be argued that it would be difficult to randomise people to microfinance as the intervention may not be desired by all; therefore measuring effectiveness in those who did not desire it to begin with, may be problematic. Whilst a cluster RCT might guard against selection bias, a recent study 10 highlighted the current challenge in achieving randomisation due to the widespread diffusion of microfinance in some regions of South Asia leading to difficulties in identifying unexposed control clusters. Therefore, we included non-randomised studies in this review in order to not limit the evidence considered. The non-randomised studies included dealt with self-selection bias in two main ways, using either panel data in a quasi-experimental design or propensity score matching (PSM). However, additional analysis in of one of the studies included in this review suggested that the reduction in intimate partner violence demonstrated using conventional statistical methods did not hold when PSM was used 28 .
Due to the lack of high quality randomised controlled trials in this field, the vast majority of studies included in this study were cross-sectional. As a study design, cross-sectional studies do not provide the strongest level of evidence. Analysis of quasi-experimental and panel data studies proved difficult as there is currently no universally acceptable quality assessment tool. The use of the Cochrane Risk-of-Bias tool in this instance may have introduced an over-or under-estimation of the risk of bias and, consequently, the quality assessment of the study.

PROSPERO PROSPERO
International prospective register of systematic reviews International prospective register of systematic reviews Is microfinance associated with changes in women's empowerment and childhood Is microfinance associated with changes in women's empowerment and childhood nutrition, and does this vary by geographical region? A systematic review and nutrition, and does this vary by geographical region? A systematic review and meta-analysis meta-analysis

Review question Review question
Through systematic review and, if possible, meta-analysis assess whether microfinance programmes are Through systematic review and, if possible, meta-analysis assess whether microfinance programmes are associated with changes in female empowerment and the wellbeing of women over the age of 15 years associated with changes in female empowerment and the wellbeing of women over the age of 15 years Through systematic review and, if possible, meta-analysis assess whether microfinance programmes are Through systematic review and, if possible, meta-analysis assess whether microfinance programmes are associated with changes in use of a contraception method among women of reproductive age associated with changes in use of a contraception method among women of reproductive age Through systematic review and, if possible, meta-analysis assess whether microfinance programmes are Through systematic review and, if possible, meta-analysis assess whether microfinance programmes are associated with changes in childhood nutrition and whether this varies by the sex of the child associated with changes in childhood nutrition and whether this varies by the sex of the child

Searches Searches
The An attempt will be made to access unpublished studies and dissertations through a search of grey literature An attempt will be made to access unpublished studies and dissertations through a search of grey literature through www.thesis.com. through www.thesis.com.
The search will be limited to studies carried out after 1990. The search will be limited to studies carried out after 1990.
No language restrictions will be imposed.
No language restrictions will be imposed.

Types of study to be included Types of study to be included
Cross-sectional surveys, cohort studies, controlled before-and-after studies, interrupted time series, quasi-Cross-sectional surveys, cohort studies, controlled before-and-after studies, interrupted time series, quasiexperimental studies, randomised and non-randomised control/cluster trials. experimental studies, randomised and non-randomised control/cluster trials. A data extraction form will be completed for each selected study by one researcher under the following sub-A data extraction form will be completed for each selected study by one researcher under the following subheadings; publication details, study details, nature of study, intervention and results. The data extraction forms headings; publication details, study details, nature of study, intervention and results. The data extraction forms will then be reviewed by the second researcher. will then be reviewed by the second researcher. This is to be used in further analysis and synthesis of the data. This is to be used in further analysis and synthesis of the data.

Condition or domain being studied Condition or domain being studied
Any disparities will be solved be mutual consensus between the two primary researchers. If this is not possible, Any disparities will be solved be mutual consensus between the two primary researchers. If this is not possible, the input of the third researcher will be sought. the input of the third researcher will be sought.

Risk of bias (quality) assessment Risk of bias (quality) assessment
The selected studies will be assessed for risk of bias by two researchers using the Cochrane Collaboration's The selected studies will be assessed for risk of bias by two researchers using the Cochrane Collaboration's tool for assessing risk of bias in randomised controlled trials and for quality by the Newcastle-Ottawa Quality tool for assessing risk of bias in randomised controlled trials and for quality by the Newcastle-Ottawa Quality Assessment Scale in non-randomised studies. Assessment Scale in non-randomised studies.
Any disparities will be resolved be mutual consensus between the two primary researchers. If this is not Any disparities will be resolved be mutual consensus between the two primary researchers. If this is not possible, the input of the third researcher will be sought. possible, the input of the third researcher will be sought.

Strategy for data synthesis Strategy for data synthesis
Outcome measures will be extracted from the studies and used in the meta-analyses. The studies providing an Outcome measures will be extracted from the studies and used in the meta-analyses. The studies providing an appropriate measure of effect will be weighted using a quality rating system and then stratified by quality score. appropriate measure of effect will be weighted using a quality rating system and then stratified by quality score.
A descriptive analysis will be done for studies providing quantitative outcome measures not suitable for meta-A descriptive analysis will be done for studies providing quantitative outcome measures not suitable for metaanalysis. analysis.
A fixed-effects or a random-effects model will be used in pooling of the data and a suitable method of estimating A fixed-effects or a random-effects model will be used in pooling of the data and a suitable method of estimating variance in studies will be applied. variance in studies will be applied. The summary estimate of the effect size will be done in each stratum The summary estimate of the effect size will be done in each stratum according to quality score, i.e. high, medium and low quality score, and statistical tests (I-squared) used to according to quality score, i.e. high, medium and low quality score, and statistical tests (I-squared) used to check for heterogeneity. check for heterogeneity.

Analysis of subgroups or subsets Analysis of subgroups or subsets
A sub-group analysis of the measures of effect chosen will be done according to region to detect any variations A sub-group analysis of the measures of effect chosen will be done according to region to detect any variations between regions. The three regions will be geographically specified as Sub-Saharan Africa, South Asia and between regions. The three regions will be geographically specified as Sub-Saharan Africa, South Asia and South America. The results will be presented by tables within the text of the review or if possible in forest plots South America. The results will be presented by tables within the text of the review or if possible in forest plots in the meta-analysis in the meta-analysis  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59

Conflicts of interest Conflicts of interest
None known None known

Language Language
English English

Country Country
England England

Subject index terms status Subject index terms status
Subject indexing assigned by CRD Subject indexing assigned by CRD  PROSPERO PROSPERO This information has been provided by the named contact for this review. CRD has accepted this information in good faith This information has been provided by the named contact for this review. CRD has accepted this information in good faith and registered the review in PROSPERO. CRD bears no responsibility or liability for the content of this registration and registered the review in PROSPERO. CRD bears no responsibility or liability for the content of this registration record, any associated files or external websites. record, any associated files or external websites.

METHODS
Protocol and registration 5 Indicate if a review protocol exists, if and where it can be accessed (e.g., Web address), and, if available, provide registration information including registration number.

5
Eligibility criteria 6 Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered, language, publication status) used as criteria for eligibility, giving rationale.

5
Information sources 7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and date last searched. 6 Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated.

Supplement 1
Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable, included in the meta-analysis).

7
Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators.

7
Data items 11 List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made.

6
Risk of bias in individual studies 12 Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis. 14 Additional analyses 16 Describe methods of additional analyses (e.g., sensitivity or subgroup analyses, meta-regression), if done, indicating which were pre-specified.

Study selection
17 Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally with a flow diagram. 7 [ Figure  1] Study characteristics 18 For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and provide the citations. Table 2 Risk of bias within studies 19 Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12).

-11
[ Table 2] Results of individual studies 20 For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention group (b) effect estimates and confidence intervals, ideally with a forest plot.

-14
Synthesis of results 21 Present results of each meta-analysis done, including confidence intervals and measures of consistency.

[Figure 2]
Risk of bias across studies 22 Present results of any assessment of risk of bias across studies (see Item 15).

DISCUSSION
Summary of evidence 24 Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., healthcare providers, users, and policy makers).