A Mixed-Method Review of Cash Transfers and Intimate Partner Violence in Low- and Middle-Income Countries

Abstract

There is increasing evidence that cash transfer (CT) programs decrease intimate partner violence (IPV). However, little is known about how CTs achieve this impact. We conducted a mixed-method review of studies in low- and middle-income countries (LMICs). Fourteen quantitative and eight qualitative studies met our inclusion criteria, of which eleven and five, respectively, demonstrated evidence that CTs decrease IPV. We found little support for increases in IPV, with only two studies showing overall mixed or adverse impacts. Drawing on these studies, as well as related bodies of evidence, we developed a program theory proposing three pathways through which CT could impact IPV: (a) economic security and emotional well-being, (b) intra-household conflict, and (c) women's empowerment. The economic security and well-being pathway hypothesizes decreases in IPV, while the other two pathways have ambiguous effects depending on program design features and behavioral responses to program components. Future studies should improve IPV measurement, empirical analysis of program mechanisms, and fill regional gaps. Program framing and complementary activities, including those with the ability to shift intra-household power relations are likely to be important design features for understanding how to maximize and leverage the impact of CTs for reducing IPV, and mitigating potential adverse impacts. Intimate partner violence. Domestic violence. Cash transfers. Women's empowerment.

There is increasing interest among social epidemiologists and development economists in exploring the role that cash transfers (CTs) have on intimate partner violence (IPV). Social epidemiologists have demonstrated the pervasiveness of IPV globally, with one in three women estimated to experience at least one act of physical and/or sexual violence by an intimate partner in her lifetime (Devries et al. 2013a). Development economists have invested heavily in rigorous large-scale evaluations of social protection schemes, including CTs in low- and middle-income countries (LMICs). As the body of research grows and sophistication of methodology increases, there has been a push to demonstrate the impacts of CTs on a wider range of outcomes beyond immediate program objectives related to poverty and food security, including intra-household gender dynamics, and, more recently, women's experience of IPV. Thus, the fields of epidemiology and economics have converged on the importance of understanding if CT and IPV are linked, and which behavioral mechanisms may underpin this relationship.

Theoretically, the mechanisms through which CTs affect IPV depend on the design of the CT program. At their core, CTs are economic safety nets designed to reduce poverty. Absolute resource theory and stress theory hypothesize that CTs may lead to decreases in IPV by improving a household's economic situation, thereby reducing poverty-related stressors on individuals and households (Fox et al. 2002; Ellsberg et al. 2015; Vyas and Watts 2009). Additionally, many CT programs target women as the main beneficiary, thus potentially affecting power dynamics within the household. To model these power dynamics, economists use variants of non-unitary household bargaining models in which an increase in a woman's income (either earned or unearned as with a CT), may decrease violence by improving her bargaining power within the household (Tauchen, Witte, and Long 1991; Farmer and Tiefenthaler 1997). However, variants of the bargaining model also predict that an increase in women's resources may put a woman at increased risk of IPV if men feel threatened and use violence to reassert authority in the relationship (Eswaran and Malhotra 2011). Additionally, cash and other transfers targeted at women may also put them at risk if men use violence to extract cash or resources from them (Bloch and Rao 2002).

Theories in other disciplines such as marital dependency and feminism likewise offer mixed predictions of the effect of cash on a woman's risk of experiencing IPV. Women who are economically dependent on their partner and are surrounded by institutions that promote gender inequality and male authority over female behavior may be more susceptible to violence (Vyas and Watts 2009). Thus, CTs that target women may empower them both in the home and in the community, thereby reducing their risk of IPV. At the same time, if a woman's partner feels emasculated in his role as provider, or threatened by her increased independence, he may redouble his efforts to assert authority, using violence if necessary (Heise and Garcia Moreno 2002; Hautzinger 2003). As Jewkes (2002) observes:

“An inability to meet social expectations of successful manhood can trigger a crisis of male identity. Violence against women is a means of resolving this crisis because it allows expression of power that is otherwise denied.”

Finally, many CT programs include complementary activities such as trainings and/or linkages to health or educational services, either as a part of the program or as a “conditionality” intended to influence beneficiary behaviour—components which themselves could affect IPV. For example, group-based trainings attended by women could reduce IPV by improving their knowledge, self-efficacy, and self-esteem, thus enhancing their bargaining power. Frequent interactions with other beneficiaries in the community could build women's social capital and social ties (Brody et al. 2015), or increase the social cost of men's violent behavior (Stets 1991; Van Wyk et al. 2003). Since variation in program design is large—including size, duration, and targeting of transfers, and overlay of complementary activities—implementers’ routinely make critical decisions that influence the program's potential impact on diverse beneficiary populations.

While testing and validating theoretical models is needed to better understand and predict the impact that cash may have on IPV, there are also pressing programmatic and policy reasons to better understand these relationships and how they function across contexts and populations. First, the scale and reach of CT programming globally is both large and increasing. According to the World Bank's State of Social Safety Nets (2015), 1.9 billion people worldwide are enrolled in some form of social safety net, with approximately 20 programs operating in the average developing country and CTs present in nearly every country. In addition, CTs are expanding rapidly. For example, in sub-Saharan Africa (SSA), about half of countries in the region (21) had some form of unconditional cash transfer (UCT) programming in 2010—a number that reached 40 by 2014. In addition, CTs tend to be cost-effective, both in comparison with alternative in-kind transfers, as well as in comparison with alternative forms of poverty-alleviation (Margolies and Hoddinott 2014; Gentilini 2016). Because of their scale (reaching 718 million individuals globally) and relative cost-effectiveness, small changes in how transfers are designed and delivered have the potential to influence their impact on IPV at the margin (World Bank 2015). Similarly, given the possibility of backlash and increases in IPV, it is essential that donors and implementing agencies understand these risks and work to minimize unintended harm from such programs.

Recent reviews have sought to summarize evidence on this topic; however, none have been sufficient to understand the complex relationship between CTs and women's risk of IPV (Bardasi and Garcia 2014; Bastagli et al. 2016). Some focus largely on quantitative evidence and group IPV outcomes alongside other gendered outcomes such as women's decision-making, agency, fertility, or early marriage, thus providing little understanding of the mechanisms underlying the cash/violence relationship in different contexts. Those that focus more narrowly on IPV as an outcome combine cash transfers with a range of other economic strengthening interventions from microfinance and savings schemes to the impact of women's employment on IPV, making it impossible to isolate the impact of cash alone (Krishnan et al. 2010; Gibbs, Jacobson, and Kerr Wilson 2017).

In order to fill this gap, we have conducted a mixed-method review to help inform the understanding of the causal link between CTs and IPV in LMICs. First, we review the existing body of rigorous quantitative and qualitative research linking CTs and IPV, with a focus both on mechanisms underlying the results and the implications of CT design features on the IPV outcome. Second, we build a program theory and evaluate the level of evidence existing in support of the various pathways, drawing on both the reviewed CT literature and evidence from other fields that support or refute steps along the hypothesized causal pathway. Finally, we propose program design components and factors that may be key in delivering beneficial impacts, identify research gaps, and discuss how upcoming evaluations could be tailored or modified to fill these gaps.

Methods

We conducted a scoping exercise, which comprised a rapid assessment of the known literature, hand-searched articles, as well as articles obtained from general search engines (Google scholar). Based on this initial rapid assessment, we conducted interviews via Skype with six experts (researchers and implementers) with prior experience on the intersection between CTs and IPV. These interviews helped identify key literature, working papers, and ongoing studies, and pointed to mechanisms and hypotheses that leading experts considered viable as potential pathways linking CTs and IPV.

For the formal review process, searches were conducted using the following broad criteria: “cash transfers” and “violence”, “intimate partner violence” or “domestic violence”. Searches were conducted using the following electronic databases: PubMed, Medline, Web of knowledge, Web of Science, Global Health, and Social Sciences Abstracts. No search period restriction was imposed; however, we did limit our search to documents written in English and Spanish. Articles published in peer-reviewed journals and relevant grey literature were included. We ran forward and backward citation checks among all identified articles that met the inclusion criteria.

Table 1 describes the broad inclusion and exclusion criteria for our review. We focused exclusively on LMICs and included all types of cash transfers, whether they are conditional cash transfers (CCTs), UCTs, or bundled as part of multi-sectoral or component programming, regardless of their objective (e.g., food security, entrepreneurship, or old-age pensions). We excluded two cases of lump-sum CTs that were included primarily as part of entrepreneurship and micro-credit programs (in Uganda and Burkina Faso), as they were likely to vary substantially in the mechanisms and impact pathways; however, we include these two cases as part of the discussion. We focused on the outcomes of IPV (or domestic violence), which encompasses the following: physical, sexual, emotional, and/or psychological violence, including controlling behaviors, typically experienced inside the household, regardless of the specific methodology used to collect or measure each indicator. The IPV is further defined as violence between intimate partners (e.g., marital, co-habiting or dating partners), primarily experienced by women and perpetrated by men. However, we did not exclude evidence in the opposite direction. We included evidence showing impacts on one or more combinations of IPV outcomes, including those that show different impacts by violence type. We excluded studies that only used proxy measures for IPV, including general terms such as “conflict”, “disputes”, or measures of autonomy or empowerment. For empirical studies, we focused on methodologies that allowed a credible identification of the counterfactual, typically either randomized controlled trials (RCTs) or quasi-experimental designs with data collection at two or more points in time. For qualitative studies, we used the consolidated criteria for reporting qualitative research (COREQ) checklist (Tong, Sainsbury, and Craig 2007).¹ Two independent researchers scored the articles using three domains—(a) research team and reflexivity, (b) study design and methods, and (c) data analysis and reporting—to assign a score of high, medium, or low quality. We did not exclude any studies according to this assessment but report on the scores achieved by each study.

Table 1.

Inclusion and Exclusion Criteria for Review of Cash Transfers on Intimate Partner Violence.

Domain	Inclusion criteria	Exclusion criteria
Setting	Low- and middle-income countries	High-income countries
Program typologies	CTs (regular income support over time), regardless of stated objectives of program, in addition to or alongside other complementary interventions.	Other forms of social safety nets, including micro-loans or financing, public works or employment programs; lump-sum CTs if they are a minor component of the above programming typologies.
Indicators	Emotional, physical, sexual IPV (including homicide and assault), controlling behaviors, psychological and economic violence between co-habiting, dating or marital partners.	Proxy measures for IPV such as “conflict”, “disagreements”, “disputes” or autonomy/empowerment measures, as well as perpetration from non-partners.
Methodology (quantitative)	Use of rigorous methodology to link CTs to IPV, including a credible counterfactual.	Does not provide sufficiently rigorous research design, or description of analysis to credibly claim that pattern or results can be attributed to the program.
Methodology (qualitative)	Studies that explicitly discuss and provide evidence on the link between CTs and IPV; In order to assess quality of studies, we used the COREQ checklist. These were scored on a high, medium and low scale.	Studies were assessed on the basis of methodological limitations of individual studies, relevance to the review question, coherence across studies and adequacy of data.

Source: The authors.

Note: CTs = Cash transfers; IPV = Intimate partner violence; COREQ = Consolidated criteria for reporting qualitative research.

For both quantitative and qualitative studies, we first read the content to identify themes and mechanisms. Thereafter, we developed a matrix summarizing key information regarding the program design, implementation, and features of interest. For quantitative studies, we compiled information on methodological design, sample sizes, indicators, and impacts. For qualitative studies, we summarized methods, sample sizes, and implied impact of CT and IPV (increase, decrease, mixed, or null impacts). Where available, information was also extracted for both types of study on the underlying mechanisms that authors advanced or tested as being possibly responsible for the impacts observed. Descriptions or mechanisms that relied on the interpretation and opinion of the authors were treated as theoretical insights (hypotheses), rather than evidence.

To further refine the program theory and assess different steps in the hypothesized causal chain, we conducted comprehensive but non-exhaustive reviews of other bodies of literature. We employed snowball sampling to identify additional studies for further explanation-building, such as tracking citations in footnotes, endnotes, and references of potentially relevant articles. The protocol was registered in the Prospero database (CRD42015024511).

Results

Review of Programs and Quantitative Evidence

Table 2 summarizes the program components from the identified core quantitative papers, organized alphabetically by country and by year of publication. For the quantitative evidence, we report impacts for all qualifying IPV indicators analyzed as part of the study. However, we do not present results for each sub-sample or heterogeneity analysis. Instead, we summarize results of additional analysis in column 13 to help unpack potential mechanisms.²

Table 2.

Review of Core Quantitative Papers with Impact Evaluation Evidence on Cash Transfers and Intimate Partner Violence.

			Program details				Evaluation details			Program impacts
No	Authors	Country	Program	Modality	Transfer amount	Recipient and target group	Design	Sample size	Data collection	Indicator(s)	BL mean	Effect size (s)	Additional analysis and hypothesized mechanisms
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)
1	Roy et al. (2017)†	Bangladesh	Transfer Modality Research Initiative (NGO)	Cash and food (with nutrition BCC)	∼1,500 Taka (US$19)/month equivalent	Mothers of child aged 0–24 months in poor HHs	RCT	2,231 women	2012, 2013, 2014, 2014/151	Emotional or physical IPV (6 month, transfer)Physical IPV (6 month, transfer)Emotional IPV (6 month, transfer)Emotional or physical IPV (6 month, transfer + BCC)Physical IPV (6 month, transfer + BCC)Emotional IPV (6 month, transfer + BCC)	0.67#0.27#0.63#0.67#0.27#0.63#	ME: 0.02 (0.04)ME: 0.00 (0.02)ME: 0.03 (0.04)ME: -0.04 (0.04)ME: -0.07 (0.03)**ME: -0.02 (0.04)	Cross-sectional analysis on post-endline (e.g., 6–10 months after transfers end). Suggests transfers + BCC lead to sustainable reductions in IPV through 3 primary pathways: 1) increases in women's threat points (including financial and social empowerment, 2) greater social cost to men of inflicting IPV and 3) household economic well-being. Descriptive evidence suggests transfers alone decreased physical IPV during the transfer period.
2	Perova, Reynolds and Muller (2012)†	Brazil	Bolsa Familia (Govt)	UCT and CCT (health and education)	∼US$11/month per child (cap at 3) and UCT ∼US$34/month	Mothers in HHs with per-capita earnings < $US7/month	Quasi-exp	12,103 municipalities	2000–2010	Female homicide per 100,000 women (aged 15–49)2	6.5	OLS: -0.006 (0.0032)*	Analysis shows CT increases likelihood of women finding income generating employment, suggesting CT may affect women's bargaining power. Heterogeneous analysis shows effects are stronger for women with higher than primary education, suggesting it is more economically feasible for more educated women to leave abusive relationships, or that more education is associated with greater empowerment needed to change the status quo.
3	Rodriguez (2015)†	Colombia	Familias en Accion (Govt)	CCT (health and education)	Health and educational subsidies, up to US$64, approx. 25% of legal minimum wage (for HHs with 2 children aged 6 and 12)	Mothers in extremely poor HHs based on PMT	Quasi-exp	1,061 municip- alities	2007–2010	IPV reports per 10,000 inhabitants (monthly)3	1.57	OLS: -0.0538 (0.0183)***	Author hypothesizes mechanisms of women's empowerment and poverty-related stress, however are unable to directly test pathways.
4	Hidrobo and Fernald (2013)*	Ecuador	Bono Desarrollo Humano (Govt)	UCT	US$15/month, 6% to 10% of average BL HH expenditure	Mothers (at least one pre-school age child or no children older than 6), elderly and handicapped in bottom 40% poverty quintiles	RCT	1,250 women	2003/2004–2005/2006	Physical IPV (lifetime)Emotional IPV (lifetime)Controlling behaviors (lifetime)Psychological IPV (index, lifetime)4	0.280.530.56NR	ME: -0.02 (0.03)ME: -0.02 (0.03)ME: -0.06 (0.03)**OLS: -0.09 (0.04)**	Heterogeneous analysis shows impacts are driven by mothers with > 6 years of education (decreases in emotional IPV of 8 pp and controlling behaviors of 14 pp among subsample). In addition, among mothers with < 6 years of education, when she has equal or more education than her partner, emotional violence increases (9 pp). Results suggest a bargaining mechanism.
5	Hidrobo et al. (2016)*	Ecuador	World Food Programme pilot (United Nations)	Cash, food and voucher (nutrition trainings)	US$40/month equivalent, ∼11% of HH BL monthly consumption	Women (men if no qualifying women in the HH) according to PMT designed to include Colombian refugees in 7 urban/peri-urban centers	RCT	1,226 women	2011–2011	Physical or sexual IPV (6 month)5Emotional IPV (6 month)5Controlling behaviors (6 month)5	0.20#0.35#0.36#	ME: -0.06 (0.03)**ME: -0.04 (0.03)ME: -0.07 (0.03)**	Analysis by treatment arm (cash, voucher, food) show there is no significant difference by arm. Authors provide suggestive evidence transfers work through increased female bargaining power, increased domestic labor of both husband and wife, as well as decreased poverty-related stress.
6	Haushofer and Shapiro (2016)*	Kenya	Give Directly (NGO)	UCT	US$404 (PPP) vs. US$1,525 (PPP) and monthly vs. lump sum (RS): Average transfer US$709 (PPP) equal to twice annual per capita exp.	Women and men (RS) based on HHs with thatched (vs. metal roof) as poverty proxy in Rarieda district	RCT (HH level within-cluster and between cluster)6	1,010 women	2011–2012	Physical IPV (6 month, within villages)Sexual IPV (6 month, within villages)Emotional violence (6 month, within villages)Physical IPV (6 month, across villages)Sexual IPV (6 month, across villages)Emotional violence (6 month, across villages)	0.29#0.09#0.89#0.29#0.09#0.89#	OLS: 0.01 (0.03)OLS: -0.02 (0.02)OLS: 0.04 (0.02)*OLS: -0.07 (0.04)*OLS: -0.05 (0.03)**OLS: 0.03 (0.02)	Main analysis includes IPV as part of an empowerment index. Appendix results indicate reductions driven by spillovers within T villages. Psychological well-being increased due to CT, indicating mechanism through poverty-related stress. Heterogeneous analysis by lump sum vs. monthly transfer, by large vs. small transfer and by male vs. female recipient in main sample show no differential impacts, however in the sample including within-cluster spillover HHs, impacts are driven by sample with large, monthly CTs given to females.
7	Bobonis, Gonzalez-Brenes and Castro (2013)*	Mexico	Oportuni- dades (Govt)	CCT (school attendance, health check ups and trainings)	Education subsidy: 70–625 pesos (US$62.5)/month per HH (in 1998) Health subsidy: 12 pesos/month and nutritional supplements for children aged 4–24 months, pregnant and breastfeeding women. Basic income support: 100 pesos/month per HH with total transfer ∼10% average HH exp	Women according to a local-level marginality index (community selection) and poverty-based PMT (HH selection)	Non-exp7	2,867 women	2003	Physical or sexual IPV (12 month)Physical IPV (12 months)Sexual IPV (12 months)Threat of physical IPV (12 months)Emotional IPV (12 months)	0.160.110.090.080.11	OLS: -0.096 (0.047)**OLS: -0.062 (0.032)*OLS: -0.066 (0.039)*OLS: 0.025 (0.037)OLS: 0.041 (0.040)	Heterogeneous analysis shows impacts are concentrated among women with low expected gains to marriage (as measured by women secondary schooling completion and men's education attainment). In addition, there are positive impacts on threat of violence but no physical violence (3 pp) as well as emotional IPV with no physical (4 pp), suggesting a potential trade off between physical and emotional violence. Results suggest a bargaining and empowerment pathway as well as framework where men may seek to extract rents from female partners.
8	Bobonis, Castro and Morales (2015)†	Mexico	Oportuni- dades (Govt)	Same as above (Bobonis, Gonzalez-Brenes and Castro 2013)	Same as above (Bobonis, Gonzalez-Brenes and Castro 2013)	Same as above (Bobonis, Gonzalez-Brenes and Castro 2013)	Non-exp8	4,705 women5,800 women	20062011	Physical or sexual IPV (12 month)Physical IPV (12 months)Sexual IPV (12 months)Threat of physical IPV (12 months)Emotional IPV (12 months)Physical or sexual IPV (12 month)Physical IPV (12 months)Sexual IPV (12 months)Threat of physical IPV (12 months)Emotional IPV (12 months)	0.1370.0990.0690.0420.0710.1020.0740.0420.0220.060	OLS: 0.024 (0.026)OLS: 0.011 (0.024)OLS: 0.017 (0.020)OLS: 0.012 (0.018)OLS: 0.011 (0.020)OLS: 0.002 (0.018)OLS: -0.001 (0.015)OLS: 0.000 (0.012)OLS: 0.004 (0.010)OLS: 0.005 (0.015)	Similar to Bobonis, Gonzalez-Brenes and Castro (2013), authors investigate potential of substitution (threats with no associated violence) and find little support for this dynamic. Authors hypothesize martial selection (specifically marital dissolution among violent couples) may play a role in the lack of impact on IPV, as well as diffusion of norms rejecting IPV over time, however are unable to test the latter conclusively.
9	Angelucci (2008)*	Mexico	Oportuni- dades (Govt)	CCT (school attendance, health check ups and trainings)	Same as above (Bobonis et al. 2013)	Same as above (Bobonis et al. 2013)	Cross- sectional RCT	12,700 HHs	1998	Aggressive behavior9	0.047#	OLS: -0.006 (0.005)	Heterogeneous analysis by number of eligible children in HH indicates in HHs with no children, aggressive behavior decreases (1.6 pp), however there is increasing aggressive behavior among those HHs with specific numbers of children (2, 3 and 5). Similarly, HHs receiving a larger transfer have increased risk of aggressive behavior (transfer amount and numbers of children are highly correlated) and among HHs with uneducated husbands, particularly with partners who are of younger ages. Authors hypothesize mechanism whereby husband with traditional views of gender roles has identify threatened.
10	Rivera, Hernández and Castro (2005)†	Mexico	Oportuni- dades (Govt)	CCT (school attendance, health check ups and trainings)	Same as above (Bobonis, Gonzalez-Brenes and Castro 2013)	Same as above (Bobonis, Gonzalez-Brenes and Castro 2013)	Cross- sectional RCT10	2,861 women (urban areas)	2002–2004	Any IPV (12 months, external C group)Physical IPV (12 months, external C group)Sexual IPV (12 months, external C group)Psychological IPV (12 months, external C group)Economic IPV (12 months, external C group)Any IPV (12 months, internal C group)Physical IPV (12 months, internal C group)Sexual IPV (12 months, internal C group)Psychological IPV (12 months, internal C group)Economic IPV (12 months, internal C group)	0.4040.2200.1340.3690.1140.3640.1560.0710.3400.078	OR: 1.32 (1.06–1.63) (S)OR: 1.23 (0.95–1.59)OR: 1.21 (0.89–1.66)OR: 1.33 (1.07–1.65) (S)OR: 1.17 (0.84–1.63)OR: 1.14 (0.82–1.59)OR: 0.76 (0.49 -1.16)OR: 0.63 (0.36–1.10)OR: 1.26 (0.91–1.76)OR: 0.54 (0.29–0.99)	ORs are for internal C, rather than T. Results show that in two cases, there are decreases in IPV in comparison to external C group (but not seen when comparison is internal C group). This indicates there may be positive spillover effects for non-program women in the same communities, however no mechanisms formally tested.
11	Perova (2010)†	Peru	Juntos (Govt)	CCT	100 soles, approx. 30% of the average HH monthly income	Women in HHs with at least 1 child aged < 14 years and poverty-based PMT (with community verification)	Quasi-exp11	3,904 women	2004, 2005, 2006, 2007, 2008	Physical IPV (12 months)Sexual IPV (12 months)Emotional IPV (12 months)	0.140.040.15	ME: -0.09 (0.04)**ME: -0.03 (0.03)ME: -0.11 (0.05)**	Heterogeneous analysis by number of children, exposure to violence as a child and women's availability of a cash-paying job indicates in general, effects are higher for women with better outside options (cash-paying job) and lower for those with worse marriage markets (higher number of children or exposed to violence as a child).
12	Ritter Burga (2014)†	Peru	Juntos (Govt)	CCT	Same as above (Perova, 2010)	Same as above (Perova, 2010)	Quasi- exp12	8,065 women	2005, 2006, 2007, 2008, 2009, 2010, 2011	Physical IPV (12 months)Sexual IPV (12 months)Emotional IPV (12 months)	0.14#0.04#0.14#	OLS: -0.065 (0.035)*OLS: -0.060 (0.034)*OLS: -0.017 (0.020)	Authors hypothesize female empowerment mechanism (demonstrated via impacts on female decision-making). Authors also hypothesize increased interaction with the health sector due to CT conditions could make husbands fearful of discovery, or that increases in women's health could make them more resilient or able to resist victimization.
13	Pettifor et al. (2016)*; Kilburn et al. (2018)*	South Africa	HIV Prevention Trials Network 068	CCT (school attendance)	Young women: ZAR100 (∼US$10)Parents: ZAR 200 (∼US$20)Approx. 15.7% of BL HH expenditure	Never married or pregnant young women aged 13–20 attending grades 8–11	RCT	2,533 young women	2012–2015	Physical violence from a partner (12 months)Forced sex by a partner (12 months)	0.28#0.022#	RR: 0.66 (95% CI 0.59–0.74)RR: 1.13 (95% CI 0.75–1.70)	Suggests CT may enable young women to leave or not engage in violent relationships, as transfers reduced the number of past year sexual partners as well as delayed sexual debut.
14	Borraz and Munyo (2017)†	Uruguay	Ingreso Cuidadano and Plan de Equidad (Govt)	CCT (school attendance and health check ups), no punitive sanctions	Initially UY$1,360 or US $56/month (∼50% of BL HH income), then doubled to US$110 in 2008	Mothers of children. HH poverty-based PMT	Quasi- exp13	1,656 months (68) in police jurisdictions (24) (Monte- video)	2005–2010	Domestic assault reports per 1,000 inhabitants (monthly)	7.33	OLS: -0.124 (0.046)***	Authors suggest an empowerment mechanism, however are not able to directly test the hypothesis.

Source: The authors.

Notes: Asterisk * refers to peer-reviewed journal article; † refers to working paper or technical report; # refers to endline or baseline control mean; OLS is coefficient from ordinary least squares regression; ME is marginal effect from probit regression; OR and RR are odds and risk ratios from logistic regression. Significance levels bolded if significant at the conventional level: Asterisks indicate the following: * = p < 0.1, where additional levels are ** = p < 0.05; ** = p < 0.05 (reported as in the original article).

Abbreviations: BL = baseline; BCC = Behavior change and communication; C = control or comparison group; CT = cash transfer; CCT = conditional cash transfer; Govt = Government; HH = household; IPV = intimate partner violence; NGO = non-governmental organization; PMT = proxy means test; PPp = purchasing power parity; RCT = randomized control trial; RS = randomly selected; S = Significant; T = treatment or program group; UCT = unconditional cash transfer; US$ = United States dollar; ZAR = South African Rand.

1

The baseline (2012), midline (2013) and endline (2014) did not collect information on IPV. However, the post-endline survey in 2014/15 did collect this information, therefore estimates are reflective of 6–10 months after a 24-month transfer program had ended.

2

Homicide data collected from Brazil's National Mortality Data Base (SIM), complemented by the Sistema de Informacioes Hospitalares (SIH) and coded as homicide if it is classified as such, or the cause of death is aggression.

3

The data are municipality-level reports recorded by the National Institute of Legal Medicine via the National Reference Center on Violence, thus as events are reported events to health and justice systems, they are unlikely to comprise of less severe IPV (including emotional violence). Therefore, associated treatment coefficient is the month of CCT receipt (short-term effects), leveraging variation in payments across municipalities.

4

Aggregate psychological index is a z-score constructed by averaging z-scores for 7 psychological outcomes: (1) Does not allow you to see friends or family, (2) Does not allow you to study or work, (3) Ignores you, (4) Yells at you, (5) Tells you you are worthless, (6) Threatens to leave and (7) Threatens to take children. Of these individual indicators only (1) and (2) are negative and significant when disaggregated.

5

Impact of pooled treatment also analyzed on 19 disaggregated indicators, with 6 being significant: Controlling behaviors (1, accused you of being unfaithful, 2, tried to limit contact with your family), Emotional IPV (3, humiliated or insulted you) and Physical or sexual IPV (4, pushed you or shook you or threw something at you, 5, slapped you or twisted your arm, 6, tried to choke or burn you).

6

The study design randomized at the village level, creating pure T and C comparisons (what we report as “across villages”) and a second randomization at the HH level creating T and C within villages (what we report as “within village” estimations). The main reported effects here come from appendix table 127.

7

The authors use nationally representative household survey data (National Survey on Relationships within the Household - ENDIREH, 2003) and take a number of steps to account for potential endogeneity and construction of credible control group using within-village comparisons. In addition, the authors: (a) Restrict the sample to HHs with children < 12 years at baseline, (b) women aged 25 and older, and (c) women in unions or partnerships since 1997 or earlier (who would have made their marital choices pre-program). Models are estimated using an extensive set of control variables. Reported estimates come from table 3 (page 193), column 1 with village fixed effects, individual, and household controls.

8

Similar to Bobonis, Gonzalez-Brenes, and Castro (2013), the authors use secondary data from ENDIREH meant to capture program impacts after 9 to 13 years of exposure and create similarly restricted cohorts per cross-sectional survey. We report just the estimates for 2006 and 2011, as the 2003 estimates are already included.

9

The exact questions were: “Who is (are) the individual(s) who drinks the most in this household, irrespective of the frequency?” and “While drinking, does this person (referred to the heaviest drinker) have an aggressive behavior?”

10

The authors use an imbedded module “La Encuesta sobre Violencia y Toma de Decisiones (ENVIT)”, which was integrated into a random sample of the 2004 follow-up of the experimental evaluation for the urban sample of Oportunidades, “La Encuestra de Evaluacion de Oportunidades en areas urbanas 2004”. The survey covers both “internal controls (eligible households in communities with Oportunidades who are not beneficiaries)” and “external controls (eligible households in non-Oportunidades communities)”. Analysis gives coefficients for controls, instead of for program participation.

11

The author uses repeated cross-sections from “La Encuesta Demographica y de Salud Familiar (ENDES)” survey relying on changes in Juntos roll out, restricting analysis to the 880 highest-priority districts. Analysis controls for individual and household characteristics, including the Juntos eligibility score, as well as district and year fixed effects (coefficients reported are from fully controlled models). In addition, matching estimates are presented as robustness checks, but are not reported as they are similar to those reported here.

12

The author uses repeated cross-sections from “La Encuesta Demographica y de Salud Familiar (ENDES)” survey relying on changes in Juntos roll out and construction of a comparison group using a poverty score and limiting the analysis to poor, rural eligibles. Results reported here include difference-in-difference models with individual controls and district, year fixed effects.

13

The authors use exogenous variation in the distribution of beneficiaries across police jurisdictions before and after the second stage of the CCT program, implemented in 2008. Police Department data is used, which covers the universe of domestic assault incidents recorded in Montevideo. Difference-in-difference regressions control for population, income per capita, and unemployment rates, and include police jurisdiction and month fixed effects.

In total, we identified 14 studies meeting our inclusion criteria: six are peer-reviewed journal articles, while eight are technical reports or working papers.³ In total, nine countries are represented, with multiple studies in Mexico, Ecuador, and Peru. Only three studies were conducted in settings outside Latin America (Bangladesh, Kenya, and South Africa), and in only one case (a World Food Programme pilot in Ecuador targeted at Colombian refugees) could settings qualify as humanitarian or post-conflict. Ten out of 14 studies evaluate government programs (table 2, column 3), which have been designed as CCTs typically conditional on health and education co-responsibilities; three evaluate UCTs, with several providing additional services (e.g., behavior change communication; BCC) together with in-kind or other transfers (e.g., food or food vouchers). Programs provide a mix of flat and variable transfers (according to household size and demographic composition), ranging from 6% to 50% of baseline household expenditures (table 2, column 5). The majority of programs implement some type of means-based targeting to identify extremely poor households as beneficiaries alongside demographic criteria such as the number of children of specific ages residing in the household. Additionally, nearly all programs target women as the main recipient, with the exception of one program in Kenya that randomizes targeting to women or men. Finally, the majority of programs deliver benefits on a monthly basis (table 2, column 6).

Study designs are nearly all experimental (seven are either longitudinal or cross-sectional RCTs) or quasi-experimental (five), with the remaining two using non-experimental designs (table 2, column 7). Sample sizes at the individual level range from 1,010 women (Kenya, Give Directly) to 8,065 women (Peru, Juntos). Additionally, several evaluations used administrative data aggregated typically at the municipal level (in Brazil, Colombia, and Uruguay). Data collection for studies ranges from 1998 to 2015, with most taking place from 2004 to 2012 (table 2, column 9). In only one case did authors collect data post-intervention (e.g., 6 to 10 months after program completion) to assess if impacts were sustained after the program had ended (Roy et al. 2017).

The 14 studies examine a range of IPV outcome indicators (table 2, column 10). Overall, 56 outcomes are analyzed, including 34 measures of physical or sexual violence (13 physical violence, 10 sexual, four combined physical and/or sexual, two combined physical and/or emotional violence, two combined physical/sexual/psychological and economic violence, two IPV reported to the health and justice systems, and one administrative data on homicide). Additionally, 13 studies use measures of emotional violence, and 13 use other typologies (two controlling behaviors, three psychological violence, two economic violence, three threats of physical IPV, two combined measure of physical/sexual/psychological and economic violence, and one aggressive behavior).⁴ It should be noted that some experts conceptualize controlling behaviors as a risk factor for IPV, rather than a type of violence itself. The studies operationalize IPV in a variety of ways. The majority use some form of the conflict tactics scale (CTS), with recall periods typically six to 12 months, while a minority include lifetime measures (the latter may be less sensitive to a short-term intervention). The exceptions are the three papers that used administrative data, as well as one that asked about aggressive behavior following a partner's consumption of alcohol (Rivera, Hernández, and Castro 2005). For the mean and effect size, we have maintained the same number of significant digits or reporting as in the original reviewed papers.

Across all 56 outcomes, 20 (36 percent) are statistically significant and negative at the p < 0.10 level or higher (suggesting that the CT reduced IPV), while only one (2 percent) is statistically significant and positive at the p < 0.10 level or higher (suggesting that the CT increased IPV). The remaining 63 percent show no significant change in IPV due to the CT. For significant reductions in IPV, the percentage varies by category of violence examined: 44 percent of studies assessing physical and/or sexual IPV and 38 percent assessing other outcomes (e.g., controlling behaviors) demonstrate a significant reduction in violence, whereas only 8 percent of those assessing emotional violence, do so. The one case where an increase is found in emotional IPV is in the Give Directly pilot in Western Kenya when comparing treatment to non-treatment households in the same villages (Haushofer and Shapiro 2016). However, in the Kenya evaluation, reductions are also found for both physical and sexual violence when comparing alternate study arms (e.g., what the authors term the “across village”, rather than “within village” estimates). Although we do not formally compute average effect sizes, when considering the 13 coefficients on indicators of individual (rather than administrative) impacts, decreases range from 11 percent to 66 percent reductions over baseline means (or endline comparison means). Further, nine of these impacts represent reductions of 30 percent or more, which is quite notable given that most evaluations took place over the short or medium term.

When considering study-level impacts, overall, 11 out of the 14 studies find decreases in IPV attributable to the program, one finds mixed impacts (both decreases and increases), and two find no impacts. The two studies finding no impact are both from Mexico. One of the studies looks at long-term impacts of Oportunidades approximately nine to 13 years after program initiation empirically through the creation of comparable beneficiary and non-beneficiary groups using national surveys (Bobonis, Castro, and Morales 2015). The authors hypothesize that this lack of impact, which contrasts with the decreases they find in the short term, could be due to marital dissolution and decreases in overall rates of IPV over time; however, they are unable to test these theories comprehensively. The second study examines aggressive behavior following alcohol consumption using data from the 1998 round of the experimental Oportunidades evaluation (Angelucci 2008). Although no average effect is found, there are treatment effects (both positive and negative) by certain household characteristics and by transfer size—however, these are likely to be endogenous and therefore it is unclear how these differential effects should be interpreted.

Authors have put forward various ideas about how CT programs could affect a woman's risk of violence, but few have tested their hypothesized mechanisms empirically. Cash could decrease violence by: (a) Increasing women's empowerment or bargaining power, or changing intra-household gender dynamics (mentioned by all 12 studies documenting decreases, except Rivera, Hernández and Castro 2005; with evidence from all but Rodriguez 2015 suggesting that the pathway could be valid). (b) Decreasing household poverty and therefore poverty-related stress or improving emotional well-being (mentioned by four studies: Rodriguez (2015); Hidrobo, Peterman, and Heise (2016); Haushofer and Shapiro (2016),Roy et al. (2017); with evidence from all but Rodriguez (2015) suggesting that the pathway could be valid). (c) Increasing interaction with the health sector, thereby improving women's overall health and making her more resilient to abuse (mentioned by one study, Ritter Burga 2014, including evidence suggesting the pathway could be valid). (d) Encouraging greater interaction with other women and village leaders, which increases a woman's social capital and social ties, and could increase the social cost of men perpetrating violence (mentioned by one study, however, not tested directly: Roy et al. 2017).

In only two cases do authors hypothesize reasons for potential increases in IPV, including: (a) A partner seeking to extract resources/CT from his wife (mentioned by one study: Bobonis, Gonzalez-Brenes, and Castro 2013, however not tested directly); and (b) male backlash, specifically due to partners feeling threatened by women usurping their traditional “identity” as a provider (Angelucci 2008; with evidence suggesting the pathway could be valid).

Of note, only in one study (Hidrobo and Fernald 2013) do authors acknowledge that there may be multiple mechanisms at play that could cancel each other out (e.g., female bargaining and male backlash).

Review of Programs and Qualitative Evidence

Table 3 summarizes the program components from the identified core qualitative papers, organized alphabetically by country and author and year of publication.⁵

Table 3.

Review of Core Qualitative Papers with Evidence on Cash Transfers and Intimate Partner Violence.

			Program details				Evaluation objective and findings
No.	Authors	Country	Program	Modality	Transfer amount	Recipient and target group	Design and sample size	Data collection	Study objectives	Direction of effect	Thematic results or hypothesized mechanisms	COREQ assessment
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
1	Buller et al. (2016)*	Ecuador	World Food Programme pilot (United Nations)	Cash, food or voucher + nutrition trainings	US$40/month (equivalent in food or vouchers), ∼11% of HHs BL monthly consumption	Women (men if no qualifying women in the HH) according to PMT designed to include Colombian refugees in 7 urban/peri-urban centers	48 IDIs and 2 FGDs with women randomly selected from quantitative sample stratified on changes in IPV and transfer arm; 6 FGDs with men randomly selected from quantitative sample stratified on whether partner or male received transfer	2013	1) Explore women's experiences of transfers, household and relationships dynamics and experience of IPV; 2) Understand how transfers might disrupt gender relations and affect IPV; 3) Explore how men interpreted and reacted to introduction of transfers	Decreased	Transfers led to reductions in IPV through 3 pathways: 1) Reduced day to day conflict and stress in the couple by improving financial means to fulfil basic needs; 2) Improved HH well-being and happiness; 3) Increased women's decision making, self-confidence and freedom of movement (potentially related to improved financial situation, increased nutritional knowledge, and/or increased participation in social groups through attendance at nutrition workshops). Transfers were not perceived as threatening to men because traditional gender norms dictate shopping and food preparation are within the “woman's domain”.	High
2	Slater and Mphale (2008)†	Lesotho	World Vision pilot (NGO)	UCT only, food only and cash + food to HHs	M80 per capita (approx.US$5) cash; food transfer ∼similar value, depending on location	Drought-affected households in six constituencies in two districts	25 IDIs and 25 FGDs with beneficiaries and non-beneficiaries. FGD participants were identified by the local chief and IDI participants from within FGD participants	2007- 2008	1) What impact has the receipt of food aid or cash had on gender relations within recipient HHs? 2) Has it eased conflict over scarce resources, or created conflict over how to use or spend food aid or cash? 3) How does this compare to HHs in similar economic circumstances who have not received food or cash? 4) How are decisions being made about how widely to share cash or food within and between HHs?	Decreased	Reduction in gender conflicts and tensions (reconfirmed with author that the use of the expression ‘conflict’ included IPV) from receiving the transfers. This is because women have the primary responsibility for meeting HH allocation and budgeting needs. Further, since the amount of cash was sufficient, but not large, the scope of expenditure decisions and disagreements was relatively limited.	Low
3	Maldonado, Nájera and Segovia (2005)†	Mexico	Oportunidades (Govt)	CCT (school attendance, health checkups and trainings)	Education subsidy: 70–625 pesos (US$ 62.5)/month per HH (in 1998) Health subsidy: 12 pesos/month and nutritional supplements for children 4–24 months, pregnant and breastfeeding women. Basic income support: 100 pesos/month per HH with total transfer ∼10% average HH exp	Women according to a local-level marginality index (community selection) and poverty-based PMT (HH selection)	3 FGDs, referred to as reflexive focus groups (semi-structured), ∼10 men per group (n = 29)	Sept-Oct 2005	Explores the effect of the introduction of resources on couple and family relationships and if there is change in these relations. If there is change, then what is the type and scope of change?	No clear effect	No direct effect of cash on increase in IPV; money from the program was kept aside as money for the HH (for children's education) and not as money for the woman (although she administered it). Hence, men did not feel threatened, and there was no increase in IPV. If IPV did occur, it was related more to the dynamics of the relationship and the distribution of power within the couple prior to the receipt of transfers. In particular, it resulted from situations where violence was seen as normal by men. The authors hypothesize IPV might decrease as a product of less tension in the HH due to less money problems, however, more research needs to be done to establish this direct link.	Medium
4	Adato et al. (2000)†	Mexico	Oportunidades (Govt)	CCT (on school attendance, health checkups and trainings)	Same as above (Maldonado, Nájera and Segovia 2005)	Same as above (Maldonado, Nájera and Segovia 2005)	23 FGDS in 6 states (8 communities), with 230 participants: 80 beneficiaries, 80 non-beneficiaries and Promotoras (community workers) representing 70 communities from 7 regions	1999	Explores perceptions of women as primary recipients of transfers, women and men's attitudes toward this aspect of program design, and the positive and negative effects of this choice.	No clear effect	IPV does not seem to have been reported or freely discussed. Qualitatively, there appears to be no evidence that the receipt of cash increased IPV, but if HHs already experienced disharmony, then program requirements might have exacerbated it. Further, IPV was discussed more generally; women would not say that they experienced it themselves, but rather that they knew of such a case in their community. IPV stories were rare, either because women were uncomfortable talking about it or better research methods are needed to collect it.	High
5	Adato et al. (2004)†	Nicaragua	Red de la Proteccion Social (Govt)	CCT (on school attendance, health education workshops, child immunizations)	About US$30/month, or more depending on number of school-aged children	Women (unless single fathers) according to a stratified random sample at the locality level, Selection of localities using poverty index	SSIs, case studies and participant observations among 20 HHs in each of six communities selected on basis of social stratification to represent cross-section of community	Aug-Dec 2003	Explore the effect of the program on individuals and communities, and mechanisms for these effects.	Decreased	Few references were made by women regarding men's negative attitudes towards women that participated in the program. In general, there was an improvement in HH relations with an increase in resources due to reduced financial strain. Most women mentioned that men expressed agreement with the program selecting women as recipients, as they were more like to spend resources on the family. There were few instances where women handed over their money to men or that the beneficiary was the man. Rare instances of men forcibly taking money from wives.	High
6	Yildirim, Ozdemir and Sezgin (2014)*	Turkey	Social Risk Mitigation Project of the General Directorate of Social Assistance financially supported by the World Bank (Govt)	CCT and in-kind (school attendance, health checkups and vaccines and in-hospital deliveries)	Primary school boys US$15, girls US$20; secondary school boys US$25, girls US$30; health support US$15/month for children 0–6 years; health support for pregnant women	Children (boys and girls) of targeted group receive educational transfers based on school attendance (although transfers are higher for girls to encourage female education) and health transfers conditional on regular health center visits and vaccinations	397 IDIs with beneficiaries and key informants	2011	Explores the views, experiences, and perceptions of CT program beneficiaries.	Decreased	When beneficiaries were asked if they had experienced IPV in HHs before receipt of CCTs and whether there had been a decrease after receipt, 8% reported no IPV experienced in the HH. A majority of fights and continued IPV appeared to be due to financial difficulties. Almost 71% of victims of IPV suggested that it was decreased after they received stipends.	High
7	Angeles (2012)†	Uganda	Action against Hunger (NGO)	UCT (but also included is the form of VLSA groups, livelihoods training and GBV prevention activities)	US$170 in 2 installments	Vulnerable women in targeted areas in the north	10 FGDs with primary beneficiaries and 16 IDIs among beneficiaries and staff	2011	Explores how the CT intervention reduced GBV and how the program worked.	Decreased	According to women, transfers helped reduced financial stress. There was a decrease in fights that occurred within couples as they could avoid fighting over scarce resources and competing over meagre resources. The CT helped pay for a number of items such as school fees, medical bills or immediate needs.	Low
											Some women also felt that transfers increased women's financial autonomy. They were less dependent on their husbands than before on expenditures related to their personal needs. If there was not enough money to go around, fighting ensued. Women reported that ‘fighting’ meant verbal or physical violence.
8	Nuwakora (2014)†	Uganda	Action against Hunger (NGO)	UCT (but also included is the form of VLSA groups, livelihoods training and GBV prevention activities)	Same as above (Angeles 2012)	Vulnerable women in targeted areas in the north	12 FGDs, 19 IDIs and consultative meetings with stakeholders.	2014	Explores how the CT intervention reduced GBV and how the program worked.	Mixed evidence, in most cases decreased and increase in some isolated cases.	The CTs resulted in: 1) Overall reductions in physical, sexual and emotional violence; for the male members of beneficiary households, the CT presented a double-edged sword. On the one hand, it improved the financial conditions of the HH whilst relieving them of the burden of responsibility. Hence in some cases, men felt it contributed to or normalized negative masculinities; and 2) Reductions in levels of alcoholism among men. However in some isolated cases, there was an increase in IPV. Some women reported men wanting the money to drink alcohol resulting in the woman getting beaten. Also some men viewed women as over-empowered and thus men perceived women as “threats”.	Medium

Source: The authors

Notes: Asterisk * refers to peer-reviewed journal article; † refers to working paper or technical report; (a) Direction of effect demonstrates whether the CT increased, decreased, had mixed effects or had no effect on IPV; (b) The COREQ assessment is a 32-item checklist to help researchers report important aspects of qualitative research, such as the research team and reflexivity, study design and methods, data analysis and reporting. Scoring based on assessment by two independent researchers; (c) Outcomes included broader forms of GBV (e.g., sexual violence, theft).

Abbreviations: BL = baseline; CT = cash transfer; CCT = conditional cash transfer; FGD = focus group discussions; GBV = gender-based violence; Govt = Government; HH = household; IDI = in-depth interview; IPV = intimate partner violence; M = Lesotho Maloti; NGO = non-governmental organisation; PMT = proxy means test; SSI = semi-structured interview; UCT = unconditional cash transfer; US$ = United States dollar; VLSA = village loans savings association.

In total, we identified eight qualitative studies meeting our inclusion criteria: two are published in peer-reviewed journals and seven are working papers or technical reports. In terms of quality assessment using the COREQ checklist (table 3, column 13), four of the included studies are given a high score and four are given medium scores. Overall, the studies represent six countries, including two assessing Oportunidades/Progresa in Mexico, one each from Ecuador and Nicaragua, two from SSA (Uganda, and Lesotho), and one from Turkey (table 3, column 2). Three of the qualitative studies are NGO-led programs, of which two are external evaluations of the same CT implemented by Action Against Hunger in northern Uganda in 2012 and 2014; three are government-run programs (two UCTs and one CCT); and two are run by international organizations (table 3, column 3). Of the eight qualitative studies, three interventions are UCTs: one provides cash, food, or vouchers conditional on attending nutrition training, while four are CCTs (table 3, column 4). Women are targeted as the main recipient in most programs, despite cases where the household or a small proportion of males receive the transfer (Lesotho and Ecuador; table 3, column 6). In almost all the studies, either focus group discussions (two studies), in-depth interviews (two studies) or a combination of the two methods (four studies) were used as the method of data collection. One study in Nicaragua used an ethnographic approach, with semi-structured interviews and participant observation to explore perceptions of the program (table 3, column 7). Data collection for the studies range from 1999 to 2014, with the majority taking place between 2011 and 2014 (table 3, column 8).

The eight qualitative studies explore a range of dynamics relevant to CTs and IPV, including the following: addressing how the receipt of cash has influenced household gender relations; whether conflict over resources within the household has increased or decreased; whether there has been a change in couple and/or family relationships; and whether receipt of the transfer has affected women's decision-making authority. Some studies focus specifically on these themes, while others are more general, exploring the impact of CT on poverty alleviation with sub-objectives that focus on gender relations and household decision-making (table 3, column 10). Five of the studies show a reduction in IPV after receipt of the CT (Adato et al. 2004; Slater and Mphale 2008; Angeles 2012; Yildirim, Ozdemir, and Sezgin 2014; Buller et al. 2016;), while one study shows mixed results with an overall reduction in all forms of IPV but also some isolated households where IPV increased (Nuwakora 2014). Two studies show no clear effect of the CT on IPV (Adato et al. 2000; Maldonado, Nájera, and Segovia 2005). In one of these studies the authors note that IPV was not reported freely given the sensitive nature of the topic, which might have influenced results (Adato et al. 2000). In the other study with no clear effects, Maldonado, Nájera, and Segovia (2005), who explore the impact of Oportunidades on intra-household dynamics in Mexico, hypothesize that the dedicated use of the transfer for children's education—as opposed to money for a woman herself—has meant that men have not felt threatened by the transfer, resulting in null effects on IPV.

Authors of these qualitative studies suggest that the following mechanisms could explain decreases in IPV: (a) Reductions in poverty-related stress (mentioned by five studies: Adato et al. (2004); Angeles (2012); Yildirim, Ozdemir, and Sezgin (2014); Nuwakora (2014); Buller et al. (2016), with evidence by all suggesting that the pathway could be valid). (b) Reduction in household tensions leading to fewer conflicts (mentioned by four studies: Slater and Mphale (2008); Angeles (2012); Yildirim, Ozdemir, and Sezgin (2014); Buller et al. (2016); with evidence by all suggesting that the pathway could be valid). (c) Increased women's decision-making power in the household and feelings of empowerment (mentioned by four studies: Slater and Mphale (2008); Angeles (2012); Nuwakora (2014); Buller et al. (2016), with evidence by all suggesting that the pathway could be valid).

In the few studies that mention increases in IPV, authors suggest the following mechanisms could increase IPV: (a) The forced extraction of money/cash by a woman's male partner (mentioned by two studies: Nuwakora (2014) and Adato et al. (2000), with evidence from Nuwakora (2014) only suggesting that the pathway could be valid). (b) As a compensatory mechanism to re-assert authority when a man feels his masculinity is being threatened (mentioned by one study: Nuwakora (2014), with evidence suggesting that the pathway could be valid).

Program Theory for Understanding the Relationship between CTs and IPV

Our review suggests that there are three primary pathways through which CTs may affect IPV. For ease of reference, we have named these: (a) the economic security and emotional well-being pathway; (b) the intra-household conflict pathway; and (c) the women's empowerment pathway. The first pathway operates primarily through household-level mechanisms, evolving from a pure “income effect” of cash into the household (regardless of who is the primary recipient), which reduces poverty-related stress and improves emotional well-being. The second pathway works through the effect of cash on marital dynamics and conflict: increased access to cash, particularly in very poor households, can lessen conflict by reducing arguments over limited budgets and daily money needed to run the household. Alternatively, if CT funds are used for expenditures not intended to benefit all household members, for example to purchase alcohol or tobacco, cash could create new sources of marital conflict. Finally, through the third pathway, cash or complementary interventions could, if appropriately targeted, increase a woman's bargaining power, strengthen her self-worth, and potentially increase her perceived value to the household. Similar to the conflict pathway, this may have mixed effects depending on how men respond to potential shifts in resources or power dynamics. On the one hand, some men may feel threatened in situations where their wives are empowered, which can lead to a backlash and increased IPV as men attempt to reassert control and their identity as the household provider or dominant decision-maker. On the other hand, some men may accept this elevated position of women in the household and decrease IPV in order to keep her satisfied within the marriage.

Figure 1 summarises the three pathways and articulates the various steps in the hypothesized causal chain. Design elements listed on the far left—such as the size, frequency and duration of transfers, and targeting criteria including the particular vulnerability and poverty profiles of the beneficiary population and whether or not women are explicit recipients—can influence the impact of a program. We hypothesize that the specific pathways or causal mechanisms that operate in any instance may be a function of: (a) the design features of the CT itself; (b) how a woman's partner reacts to the transfer; and (c) the context of the CT program, including underlying factors such as the gender regimes, social norms, and local laws and policies. In the following sections, we explain stylized versions of each pathway, relying on a broader evidence base than the CT and IPV literature where necessary, and analyze the degree to which data from the review either supports or refutes the hypothesized pathway.

Figure 1.

Program Theory Linking Cash Transfer and Intimate Partner Violence

Source: The authors. Note: IPV = Intimate partner violence.

Economic Security and Emotional Well-Being Pathway

As CTs are primarily designed as an economic social safety net, the most generalizable pathway that results in decreases in IPV is through improved household economic security and associated decreases in household poverty (e.g., increased financial and food security; increased savings, assets, and investments; and improved financial coping strategies). These improvements, in turn, have the potential to improve emotional well-being of household members by decreasing poverty-related stress and improving mental health. This positive effect could directly lead to decreases in IPV, or work indirectly through decreased use of alcohol as a negative coping mechanism in response to poverty and financial stress.

CTs and Increased Economic Security (Decreased Poverty)

There is a large and robust body of literature across different geographical regions and program typologies showing that, in general, CTs have significant positive impacts on a range of household-level economic-security outcomes, including poverty rates, food security, household expenditure and consumption, household durable and productive assets, income-generation and labor-force participation, and savings and investments (Hidrobo et al. 2014; Bastagli et al. 2016; Natali et al. 2016; Banerjee et al. 2017; Handa et al. 2017; Handa et al. 2018; Hidrobo et al. 2018). Further, there is a growing body of literature documenting the positive local economy impacts of CTs, implying positive spillovers on non-beneficiary households in terms of economic outcomes (Taylor, Thome, and Filipski 2016). For this pathway to be effective, program design and implementation components—such as the relative size of the transfer, and the regularity and duration of benefits—are important factors in determining the magnitude of the impact of CTs.

Economic Security and Improved Emotional Well-Being

There is increasing evidence that poverty and poor mental health are linked in a two-way, reinforcing relationship. On the one hand, poverty is a risk factor for poor mental health and mental disorders though malnutrition, stress, substance abuse, social exclusion, and exposure to trauma and violence (the social causation hypothesis), while on the other hand, poor mental health increases risk of poverty due to increased health expenditures, reduced productivity, stigma, and loss of employment and earnings (the social drift hypothesis; Lund et al. 2010).

Whereas these linkages have been well explored in developed countries, only in the last few decades has the relationship been confirmed in LMICs. Lund et al. (2010) conducted a systematic review of the epidemiologic literature in LMICs to assess the relationship between poverty and common mental disorders, and found that among 115 studies reviewed, most reported positive associations between a variety of poverty measures and negative mental health outcomes (73 percent to 79 percent of studies). However, the strength of this relationship depended on the specific poverty dimensions examined. Corroborating these findings, a meta-analysis using 60 studies finds that individuals of low socio-economic status had higher odds of being depressed (Lorant et al. 2003), and a global analysis of over 139,000 individuals from 131 countries shows a positive relationship between income and emotional well-being, both within and across countries (Sacks, Stevenson, and Wolferes 2010). These findings are supported by qualitative evidence on the effects of transfers on beneficiary families across regions. For example, a woman in Ecuador reported the following:

“In my household it was like happiness, we all got along, with my children, with my husband […] in my house we were happy […] because before we did not have enough money for those things [food],” (Buller et al. 2016). Further, a Ugandan beneficiary reports that “Apart from the cash, we have been united as a group. The project has brought happiness in the family, as husbands and wives. It has also united parents to their bigger children,” (Nuwakora 2014).

Haushofer and Fehr (2014) provide insights into the psychology of poverty by summarizing evidence that suggests poverty-related stress causes negative affective states (including sadness and anger), which increase short-sighted and risk-adverse decision-making and other economic behaviors that reinforce poverty. Overall, evidence confirms a strong relationship between poverty and mental health in developing settings.

Although the relationship between poverty and poor mental health is well established, we know less about the typologies of interventions that are successful at breaking the two-way cycle. A recent review of programming concludes that, although there is good evidence that a variety of mental health programs have positive impacts on economic outcomes, overall the mental health effects of poverty-alleviation programs are inconclusive (Lund et al. 2011). However, CCTs are identified as a caveat to the latter statement. In recent years there has also been increasing evidence that UCTs have the potential to improve mental health and well-being of children, youth, and adults in recipient households. In particular, there is evidence that CTs have positive impacts on measures of happiness and life satisfaction, stress, and depression (Ozer et al. 2011; Daidone et al. 2015; Kilburn et al. 2016a; Haushofer and Shapiro 2016), as well as child cognitive and behavioral assessments, cortisol concentration biomarkers, and adolescent psychological distress (Fernald and Gunnar 2009; Baird, De Hoop, and Özler 2013; Kilburn et al. 2016b).

Emotional Well-Being and IPV

As with poverty and mental health, evidence suggests that the relationship between poor mental health and IPV victimization is bidirectional (Machisa, Christofides, and Jewkes 2017). In a recent systematic review and meta-analysis of longitudinal studies, Devries et al. (2013b) find that, for women and men, depressive symptoms were associated with recent experience of IPV and, conversely, that recent experience of IPV is associated with recent depressive symptoms (the latter for women only). A recent study including 10,178 men in six countries in Asia and the Pacific finds that depressive symptoms increase the risk of physical, sexual, and emotional IPV perpetration after adjusting for childhood exposure to violence (Fulu et al. 2013). Alongside depression, the literature also identifies anxiety and post-traumatic stress disorder and other mental health disorders as associated with IPV victimization. For example, a review of cross-sectional psychiatric morbidity and populations surveys finds associations between all mental disorders and IPV victimization in both men and women (Oram et al. 2014), and a systematic review and meta-analysis of 41 studies finds a higher risk of experiencing IPV among women with depressive disorders, anxiety disorders, and post-traumatic stress disorder (PTSD), in comparison with women without mental health disorders (Trevillion et al. 2012).

The link between poor emotional well-being (in particular situational stress) and IPV has also been documented. Several studies, including one among couples in Thailand, have demonstrated an association between current life stressors and the risk of experiencing and/or perpetrating IPV (Hoffman, Demo, and Edwards 1994; Cano and Vivian 2001). Additionally, a study among U.S. Air Force Active Duty members documents a strong effect of financial stress on the risk of perpetrating IPV among both men and women (Slep et al. 2010). There is also emerging evidence that childhood abuse or other adversities may potentiate the impact of recent stressors on risk of IPV perpetration, a hypothesis known as the “stress sensitization theory.” Among 34,653 adults in the United States, for example, the risk of perpetrating IPV among men with high current life stress was 10.1 percentage points greater among those with histories of high versus low childhood adversity scores (Roberts et al. 2011).

Economic Security, Alcohol Abuse and IPV

A final mechanism through which improved economic security may affect the risk of IPV is through reduced alcohol consumption via improved emotional well-being. Although the relationship between economic security and alcohol use is complex, many studies show that the largest burden of alcohol-related mortality and morbidity falls on populations with low socio-economic status (Jones et al. 2015). Likewise, a robust body of evidence from LMICs shows a strong and consistent association between men's use of alcohol and women's risk of IPV (Gage 2005; Foran and O'Leary 2008; Graham et al. 2008; Hindin, Kishor, and Ansara 2008; Dalal, Rahman, and Jansson 2009; Abramsky et al. 2011); one systematic review pools the results of 11 studies and finds that harmful use of alcohol is associated with a 4.6-fold increased risk of exposure to IPV compared with mild or no alcohol use (Gil-Gonzalez et al. 2006). Studies suggest that alcohol affects risk of IPV in multiple ways: as a trigger for arguments (Heise 2012); by affecting problem-solving and other cognitive abilities (Hoaken, Assaad, and Pihl 1998); by lowering inhibitions and making it easier to misinterpret verbal and non-verbal cues (Klostermann and Fals-Stewart 2006); and by playing into culturally defined scripts about how alcohol affects behavior (Quigley and Leonard 2006). While alcohol alone is neither necessary nor sufficient to cause violence, a recent review concludes that it meets all the epidemiological criteria for being considered a contributing cause of IPV (Leonard and Quigley 2017).

Intra-household Conflict Pathway

While greater financial stability may reduce IPV by improving emotional well-being, access to cash can also affect violence directly by either reducing or increasing fodder for arguments. More cash can reduce marital conflict over money, or it can increase conflict if the money is diverted to temptation goods or partners disagree on how the money is spent. In the systematic review, Vives-Cases, Gil-González, and Carrasco-Portiño (2009) find that marital conflict is significantly associated with IPV in 10 out of 11 studies identified.

Decreased Conflicts over Money

Conflicts over money have been identified by different studies in poverty contexts as a trigger for violent episodes within couples (Rabbani, Qureshi, and Rizvi 2008; Fehringer and Hindin 2014). Our review shows that CTs seem to have an impact in reducing arguments of this type. From the papers included in our review, Buller et al.’s (2016) mixed-methods analysis finds that the provision of cash to households reduces IPV, partially by eliminating the need for women to negotiate the daily cash they need to buy food for the family. During qualitative interviews post-trial, women reported that transfers meant they did not have to ask their husbands for money, which eliminated a source of conflict in the relationship. Furthermore, Angeles (2012) finds that women in Uganda reported a decrease in fights occurring due to competition over scarce resources. The CT helped pay for a number of items such as school fees, medical bills, or immediate needs, effectively reducing the arguments over money. Likewise, Yildirim, Ozdemir, and Sezgin (2014) find that, according to respondents, a majority of fights and continued IPV appeared to be due to financial difficulties, with the majority of victims reporting IPV decreases or cessation after they had started receiving the transfer. According to a respondent in Turkey:

“There had been many fights. Because children needed many things that we could not have afforded. I asked my husband and he used to say there is no money. Then I used to get upset and started to yell. We had many fights because of poverty. Not only for us, for all poor, fights come from suffering,” (Yildirim, Ozdemir, and Sezgin 2014).

Increased Conflict over Temptation Goods

It is also possible that an unintended effect of CTs could be an increase in spending on temptation goods by either men or women. This relationship has generally not been supported by the literature, although there is limited global evidence on certain types of temptation goods (e.g., gambling and prostitution as compared with consumable goods). Evans and Popova (2017) conducted a systematic review on the link between CTs (both conditional and unconditional) in LMICs and temptation goods, primarily alcohol and tobacco. The authors included 50 estimates from 19 studies and conclude that there is no systematic evidence that beneficiaries increase spending on alcohol and tobacco—a conclusion also reached by a recent analysis of seven government UCT programs in Africa (Handa et al. 2017). It is important to note that this does not mean that cash is not partially used to purchase these goods, but rather that there is no systematic difference compared to spending in non-beneficiary households.

Women's Empowerment Pathway

CTs are often hypothesized to empower women either through increasing their direct access to cash, information (through trainings), or social networks (via group activities)—all of which can enhance women's sense of empowerment. If resources are placed in the hands of a woman, her relative control of resources within the household improves, thus increasing her bargaining power and ability to negotiate her preferences. Direct receipt of cash also increases her financial autonomy and contributes to enhanced self-efficacy and confidence, potentially shifting the balance of power between the woman and her male partner.

Depending on how her partner reacts, this shift in power can either increase or decrease a woman's risk of IPV. Greater female empowerment can strengthen a woman's ability to exit an abusive relationship or at least credibly threaten to leave, which might deter her husband from using violence. Likewise, if the man's reaction is positive and accepting, risk of violence may decrease as the man comes to appreciate both his wife's competency and the added resources she brings to the household. Greater female empowerment, however, could result in more violence if a man reacts negatively to his wife's willingness to assert her preferences more forcefully. Some men may feel threatened by this shift in power and may use violence to reassert their dominance and male authority in the family.

CTs and Empowerment

Case studies support the notion that CTs can have transformational impacts on women's empowerment through improved decision-making and feelings of independence from partners (Patel, Hochfield, and Jacqueline 2012; Nuwakora, 2014; Yildirim, Ozdemir, and Sezgin 2014). As a woman from Northern Uganda reported:

“Earlier, we used to farm as a family. However, my husband would sometimes sell household items without consulting me. But now that I have my own money, I can have a say on how to spend income. Moreover, I cultivate the gardens together with my husband […]” (Nuwakora 2014). Moreover, a woman in Mexico states that “I have seen that all mothers, like indigenous women that we are, things changed a lot. I notice it because now women participate a lot, when there is an assembly, or meeting, or “plática”. They participate a lot because they have this responsibility, in order for the support [transfer] to come,” (Adato et al. 2000).

Numerous studies also show that CTs increase women's savings and income-earning opportunities, suggesting that CTs may affect women's bargaining power (Perova, Reynolds, and Muller 2012; Green et al. 2015; Natali et al. 2016). However, the broader evidence is mixed. In a recent synthesis of qualitative and quantitative reviews and key evidence, van den Bold, Quisumbing, and Gillespie (2013) find that, although qualitative evidence on CCTs largely from Latin America and the Caribbean generally points to positive impacts on empowerment indicators, quantitative results are mixed. More recent studies focusing on the Africa region come to the same broad conclusions (Bonilla et al. 2017), and others raise competing arguments that CTs can reinforce traditional gender norms, or place an additional burden on women's time use, further reinforcing gender inequities (Molyneux 2006; Chant 2008).

At least part of the ambiguity around this linkage can be attributed to the diverse set of indicators used to measure empowerment and the inherent difficulty in drawing conclusions based on few quantitative indicators of intra-household bargaining (Peterman et al. 2015; Seymour and Peterman 2017). Adding to the complexities, intra-household empowerment is highly contextual, and there has been no clear consensus within and across disciplines of how to best measure it (Malhotra and Schuler 2005). Thus, although there are promising case studies, there are also mixed impacts, and a lack of consolidated evidence across program typologies and diverse contexts with differing gendered norms.

Shifts in Relationship Power and IPV

Another strand of literature reviews how empowerment and shifts in relationship power may decrease or increase IPV (Perova, Reynolds, and Muller 2012; Hughes et al. 2015). A woman's risk of IPV based on the extent of her financial independence and self-confidence is complex, context-specific, and contingent on factors such as socio-cultural contexts of households, characteristics of households and individuals, and particularities of empowerment processes themselves (Hughes et al. 2015). In terms of socio-cultural factors, in patriarchal contexts women's empowerment is more likely to lead to increased conflict and IPV, at least in the short term. Hence, the relative status of women and men in terms of decision-making and how their power and resources compare to each other is an important contributing factor for increased IPV (Hughes et al. 2015). This seems especially common in situations where a man is unable to fulfill his gender-ascribed role as “bread-winner” and a woman is beginning to contribute relatively more to family maintenance, or where a woman takes a job that defies prevailing social convention (Hughes et al. 2015). This aligns with research by Maldonado, Nájera, and Segovia (2005) from Mexico that shows that significant income increases to women may threaten men's status, causing husbands with more traditional gender views to reassert control through violence. Overall, however, the risk of increased IPV could also decline over time as both men's individual attitudes and broader social attitudes become more accepting of women's increased economic activity and financial autonomy (Ahmed 2005). For example, some participants in the South African IMAGE intervention reported that the increased self-confidence, social support, and communication skills gained from being part of a combined micro-finance and gender training initiative resulted in improved partner communication, preventing any conflict escalating into violence (Kim et al. 2007).

Conclusion and Policy, Programmatic and Research Implications

We conducted a mixed-method review of the impact of CTs on IPV in LMICs and have built a program theory to help understand the mechanisms behind this impact. In total, we identified 14 quantitative and eight qualitative studies that met our inclusion criteria, of which 11 and six, respectively, support the hypothesis that CTs decrease IPV. We find little support for increases in IPV, and only two of our reviewed studies had overall mixed or adverse impacts.

These findings, paired with the scale and relative cost-effectiveness of CTs, suggest that they have the potential to decrease IPV at the margin across large populations of vulnerable groups. However, across the 56 quantitative outcomes measured, approximately 63 percent are insignificant, suggesting that CTs may have different impacts on different types of violence within the same study. Transfers appear to reduce physical and/or sexual IPV more consistently than emotional abuse or controlling behaviors. This finding is an apparent contradiction, since several of the pathways focus on emotional states, which would suggest initial impacts on emotional and psychological IPV before affecting physical and sexual IPV. However, we conjecture that this could be due in part to measurement issues, as emotional IPV is measured less in studies, and with greater variability. Further, definitions of emotionally abusive acts vary across cultures and are thus more difficult to define (Garcia Moreno et al. 2004).

As CTs are primarily a policy tool to respond to poverty and vulnerability, it is unlikely that large-scale institutional programming will be designed with the specific objective of decreasing IPV. However, if small design changes have the potential to decrease IPV—a key indicator of well-being and gender equity—transfer programs have the scope to realize significant gains across sectors, at a lower cost than violence-specific programming. Research to better understand how CTs affect IPV, and under what conditions, can help policy-makers maximize these gains while minimizing any unintended negative impacts of CT programs. As the collection of IPV measures in multi-topic surveys are likely to imply significant survey logistical costs, expanding the feasibility of experimental “light touch” methods are likely to aid understanding of the dynamics in generalized programming (Peterman et al. 2017a).

We found evidence to support all three hypothesized pathways: economic security and emotional well-being; intra-household conflict; and women's empowerment. We also found substantial evidence from related literature to support each step in the proposed causal chains, with the exception of increasing violence by exacerbating conflict over the consumption of temptation goods. According to our program theory, the economic security and emotional well-being pathway is the only one that exclusively reduces IPV; the other two pathways may increase or decrease IPV, depending on whether additional cash aggravates or soothes relationship conflict and/or how men respond to women's increased empowerment. How these pathways play out depends on intra-household gender dynamics, which are, in turn, affected by local gender regimes and socio-economic inequalities within a setting or beneficiary population. Thus far, quantitative evaluations have not been well designed to measure these mechanisms, particularly those relating to relationship dynamics and behavioral intra-household measures.

The qualitative studies suggest that in highly patriarchal settings, shifts in household dynamics that are less challenging to traditional gender norms are less likely to prompt violence. Likewise, programs that generate smaller shifts in relationship power appear more easily accepted by men than those catalyzing larger disruptions (Maldonado, Nájera, and Segovia 2005; Slater and Mphale 2008). For example, Buller et al. (2016) note that increased cash and in-kind transfers to women have been accepted by Colombian and Ecuadorian men in part because they are intended for children's nutrition, a domain already within the domestic responsibilities of women. Indeed, how a program is “framed” and the meaning imbued to cash by a program's stated intent (e.g., for women's entrepreneurship versus child health) may influence the transfer's impact on gender dynamics and IPV as much as any other program feature. More “acceptable” shifts might also be achieved by making smaller, more regular transfers (conducive to small household purchases managed by women), rather than larger or lump-sum transfers. It should be noted, however, that the Kenya Give Directly study tested lump-sum versus periodic transfers and finds that the difference did not significantly affect the magnitude or impacts on IPV (Haushofer and Shapiro 2016). Understanding the importance of transfer size and other design features on intra-household dynamics is important, as economic security and poverty impacts are likely to be larger with increasing size of the transfer relative to pre-program household consumption, thus suggesting a potential program design trade-off.

The recipient of the CT is also likely to be a key factor in understanding potential impacts on IPV. While empirical evidence is scarce and mixed in terms of the impact of recipient sex on economic and human capital outcomes of transfers, there is even less evidence for how different targeting schemes affect IPV outcomes (Yoong, Rabinovich, and Diepeveen 2012). Across the studies reviewed here, the majority of CTs transfer cash to women; therefore, a large gap in knowledge remains with respect to impacts on IPV when men are the main recipient, as is the case in many programs in Africa. Haushofer and Shapiro (2016) published the only study that randomly compares male and female beneficiaries, and the authors find no differential impact on IPV. These differences will be particularly important in settings where men are the de facto recipient due to gendered mobility constraints and lower perceived cultural acceptability of transferring benefits to women (e.g., Middle East and parts of South Asia).

Lastly, the associated benefits from complementary activities such as trainings and group meetings are also likely to be a key factor that shapes how a CT program impacts IPV. Complementary activities could independently decrease IPV by empowering women through increased knowledge, which leads to increased self-esteem, social interaction, and social capital. Most CT programs reviewed are linked to some complementary activities. While the literature acknowledges that complementary activities might play a role in generating impacts, this mechanism is seldom explored explicitly. The Bangladesh study by Roy et al. (2017) is the only one that attempts to separately evaluate the impact of the transfer versus the transfer-plus-auxiliary activities. These authors find that decreases in IPV six months post-program exist only in the CT-plus-BCC group, and not in the CT-only group.

It is worth mentioning that although average impacts of the studies reviewed overwhelmingly show decreases in IPV, several studies find increases for select IPV outcomes within particular sub-groups of beneficiaries (e.g., Bobonis et al. 2013; Hidrobo and Fernald 2013). In addition, we excluded two studies where the cash transfers are one-time lump-sum grants as part of larger micro-enterprise programs with couples therapy or bundled livelihood, savings, and coaching programs. In the first study, Green et al. (2015) find that women in Northern Uganda receiving the micro-enterprise training alone have experienced increased marital control, while those with added couples therapy have not (with no impacts among either group on physical or emotional abuse). In the second study, Ismayilova et al. (2017) find that women in Burkina Faso benefiting from both arms of bundled savings and livelihoods programming have experienced reduced emotional violence; however, this effect is larger amongst those women receiving family coaching. Therefore, while our assessment is optimistic about the direction and level of impacts on IPV, we recognize that diverse programming variations are yet to be widely tested and understood.

Our review has a number of limitations. We exclude studies that explore the impact of transfers on other types of violence that may have implications for IPV, including community-level violence or intra-household violence perpetrated or directed at other household members. For example, there is increasing interest and some potential for social safety nets, including CTs, to decrease violence against children, although the evidence is weak for most types of childhood violence apart from sexual violence and abuse among adolescent girls (Peterman et al. 2017b). Conclusions around promising mechanisms for reduction of violence against children relate to several of those that we identify, including increases in economic security and decreases in poverty-related stress. This suggests the potential for CTs to affect multiple types of intra-household violence simultaneously, but no study to date has explored this potential. Likewise, transfers could decrease community violence through positive economic spillovers into non-beneficiary households, or could increase violence due to social tensions and jealousy triggered by the CT (Adato 2000; Slater and Mphale 2008; Wasilkowska 2012; Beasley, Morris, and Vitali 2016). Finally, we cannot generalize our findings on household dynamics to high-income countries or from CTs to broader social protection or economic strengthening programs.

Our findings, however, have important implications for future research. First, evaluations should carefully consider the IPV metrics to be included to ensure that they capture internationally validated measures of IPV that are sensitive to program impact (Heise and Hossain 2017). To date, we know little about how CTs may affect the frequency and severity of IPV, which would aid our understanding of dynamics at the margin. Second, studies need to go beyond impact to include validated and credible measurement of pathways to better understand the behavioral underpinnings of the CT and IPV relationship. In doing so, studies will deepen both our understanding of how transfers affect IPV, and our understanding of the behavioral relationships beneath each causal link, many of which are understudied in LMICs. It is likely that mixed-method studies will advance our understanding of these links better than either quantitative or qualitative studies alone; however, to date, few mixed-methods evaluations have been conducted.

There is also a need for a better understanding of how program design features affect ultimate outcomes and pathways, particularly with respect to targeting, complementary programming, program linkages, and conditionalities. Of the quantitative studies included in the review, only four use a research design that is able to test program variations (Green et al. 2015; Haushofer and Shapiro 2016; Hidrobo et al. 2016; Roy et al. 2017), and none were able to test potential synergistic effects between program components.

There are large regional and contextual gaps in our understanding of dynamics, with evidence skewed to Latin America and little understanding of Asia and the Middle East, or of how dynamics may differ in humanitarian settings. Evidence from SSA is scarce (particularly empirical evidence) and is concentrated in Eastern and Southern Africa, with little evidence arising from Western and Central Africa, where gender norms and institutions may vary. Finally, we know little about long-term impacts, including how impacts may vary over time horizons and if impacts are non-linear, as well as the sustainability of impacts after CTs end or households graduate (the latter was studied only by Roy et al. 2017).

Although our review indicates that CTs are promising tools to reduce IPV, this relationship is complex and there are large gaps in our understanding of what program design components are necessary or beneficial in diverse settings. For example, it is likely that within any one program there are multiple or competing casual pathways in operation, with differential distributional impacts or those that vary by type of IPV. It is also possible that impacts in the short run may differ from longer-term impacts as relationships begin and end and programs are phased out. Although we have not conducted a meta-analysis due to variation in outcomes captured, as the evidence base grows, future work may be able to capture variation in magnitude of impact and how it relates to key program design features, including transfer size and important contextual factors such as baseline prevalence of IPV. As cash and other transfers are increasingly scaled up in development settings, we welcome further research to better understand and leverage gains across sectors on non-traditional outcomes including IPV.

Notes

Ana Maria Buller is an Assistant Professor in the Department of Global Health & Development, 15-17 Tavistock Place, London School of Hygiene and Tropical Medicine, WC1H 9SH. email: [email protected]. Amber Peterman is a Social Policy Specialist in the Social and Economic Policy Unit, Piazza S.S. Annunziata 12, UNICEF Office of Research--Innocenti, 50122. Meghna Ranganathan is an Assistant Professor in the Department of Global Health & Development, 15-17 Tavistock Place, London School of Hygiene and Tropical Medicine, WC1H 9SH; email: [email protected]. Alexandra Bleile is a Research Assistant in the Department of Global Health & Development,15-17 Tavistock Place, London School of Hygiene and Tropical Medicine, WC1H 9SH; email: [email protected]. Melissa Hidrobo is a Research Fellow in the Poverty, Health, and Nutrition Division at the International Food Policy Research Institute, Dakar, Senegal; email: [email protected]. Lori Heise is a Professor in the Department of Population, Family and Sexual and Reproductive Health at the Johns Hopkins Bloomberg School of Public Health, with a joint appointment at the School of Nursing, 615 N. Wolfe Street, Baltimore, Maryland 21205. Email: [email protected].

This work was supported by an anonymous donor. Salary support for Peterman was provided by the UK Department for International Development (DFID). The authors thank Giulia Ferrari, Johannes Haushofer, Deborah Hines, Mazeda Hossain, Michael Naranjo, Elizaveta Perova, Audrey Pettifor, Jeremy Shapiro, Jo Spangaro, Sedona Sweeney, Seema Vyas, and Geoff Wong for helpful discussions at project inception. Tia Palermo, participants at the Sexual Violence Research Initiative Forum in September 2017, and three anonymous reviewers provided helpful comments on an earlier draft.

Footnotes

1

The COREQ checklist is a 32-item list for interviews and focus group discussions to assess the quality of the qualitative studies that have been included in this review (Tong, Sainsbury, and Craig 2007).

2

The justification for this decision is primarily because studies vary to the extent that they conduct additional analysis, which would potentially skew results towards specific studies. In addition, the choice of which sub-sample analysis is explored is left to authors; thus, we may not capture an unbiased picture of potential heterogeneous effects when comparing results side by side. Finally, the methodology to analyze sub-samples varies by study. It should be noted that no study was explicitly set up (sampled) to conduct heterogeneity analysis.

3

We count Pettifor et al. (2016) and Kilburn et al. (2018) in South Africa as one study, as the latter extends the former to examine pathways and additional indicators but fundamentally examines the same sample using the same methodology. In contrast, while other papers examine the same program (Oportunidades, Juntos), they have utilized different samples, rounds of data or use different methodology—thus, we count them as separate studies.

4

Note that these counts by violence typology sum to greater than 56, as certain classifications count in several categories for combined indicators.

5

Some of the studies use mixed methods, but the quantitative sections do not meet the inclusion criteria for the quantitative part of this review. We have only analyzed the qualitative data and, to the extent possible, have presented results from the qualitative sections.