Theoretical Underpinnings and Meta-analysis of the Effects of Cash Transfers on Intimate Partner Violence in Low- and Middle-Income Countries

Abstract

The number of studies examining the effects of cash transfer (CT) programs on Intimate Partner Violence (IPV) has rapidly grown over the last decade. Depending on how violence is modelled, CTs could either increase, decrease or have an ambiguous effect on violence. This paper provides a survey of the theoretical and quantitative empirical literature on the effects of CTs on IPV. We place the existing theories in the context of an overarching model of household bargaining. We then review the empirical evidence for low- and middle-income countries. The bulk of the empirical evidence suggests that CTs either are associated with a decrease in IPV or no effect on average. Some studies however report increases in IPV for some subgroups, for example, for women with low levels of education whose husbands have even lower levels of education. A meta-analysis finds significant negative impacts on physical and emotional violence and controlling behaviours – consistent with household resource and stress theory, possibly in conjunction with, but dominating, theories of status inconsistency and instrumental violence.

Acknowledgments

This paper was prepared with research assistance provided by Zan Fairweather. We thank two anonymous referees for very helpful comments.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. Intimate Partner Violence refers to any violence between partners in an intimate relationship. Domestic violence is often used synonymously with IPV but can also encompass child or elder abuse or abuse by any member of a household, which is beyond the scope of this paper.

2. This is as opposed to assuming two individuals who form a household pool income and maximise a household utility function, subject to standard income and time constraints as in early neoclassical models of unitary household decision-making that stem from Becker (1965). These unitary household models have been superseded by household bargaining models which better explain the experimental evidence (Iyengar & Ferrari, 2011).

3. The concept of utility in economics represents the well-being attained by an individual as a result of a set of choices.

4. We refer to the relationship as a marriage but the model applies also to other types of intimate cohabitating relationships. Cohabitation outside marriage may incur differential costs of leaving the relationship than marriage. These costs may be lower (as there are lesser legal barriers to leaving) or higher (as legal rights to the assets accruing to the relationship may be uncertain).

5. The market input (q) and the female labour used for home production ($d^f$) enter utilities only through Q. The price of the market input is normalised to 1.

6. The woman’s bargaining weight is particularly salient in cases where CTs are targeted at women and in cases where they can take the transfer with them if they decide to leave the relationship.

7. Violence can also be modelled as an outcome of loss-of-control, arising because of a negative emotional cue/shock (Card & Dahl, 2011). A distinctive feature of this model is that it occurs because individuals are boundedly rational, and emotional shocks or cues can lead to unintended/impulsive violence. A key difference between expressive violence in a household bargaining model versus a loss-of-control model is that the victim is financially compensated for violence in the former but not the latter.

8. When the woman’s threat point or reservation utility is very low (that is non-binding), for example when a woman values a marriage intrinsically or where divorce is not a credible threat in contexts where women cannot easily exit a marriage, men may use violence to capture women’s resources (Green, Blattman, Jamison, & Annan, 2015). Stevenson and Wolfers (2006) argue for instance that unilateral divorce law may reduce domestic violence by increasing the credibility of a threat point.

9. Instrumental violence has also been referred to as extractive violence (Bloch & Rao, 2002). However, instrumental violence is a broader term that captures motives used to extract resources as well as to control behaviour.

10. Tauchen et al. (1991)’s model of domestic violence includes an instrumental violence element. It sees the male’s utility as depending on some of the behaviours of his partner. These behaviours might cover a number of aspects of the relationship including the use of financial funds, contact with friends and other family members, and dress or sexual activities.

11. Additionally, the model in Haushofer et al. (2019) does not allow marital transfers, ruling out the possibility that husbands compensate wives for violence. Similarly, Perova (2010) models instrumental violence as a way for the husband to get money or control his wife’s behaviours, with violence being inherently distasteful.

12. When the woman’s reservation utility constraint is binding, violence does not really have an instrumental function since she is being held at her reservation utility level, Tauchen et al. (1991).

13. Anderberg and Rainer (2013) introduce a particular form of domestic abuse into the bargaining framework: men may use economic abuse or sabotage ‘instrumentally’ to control their partners’ behaviour by thwarting their partners’ training or career efforts. In relation to cash transfers, some men may sabotage the receipt of the money or women’s participation in information sessions, as an attempt to limit their wives’ economic involvement and opportunities.

14. As the woman preserves the transfer even if she leaves the household, the in-kind transfer ($t_k$) or the cash transfer ($t_c$) increases her reservation utility. Her relative potential income if she leaves the household is now $\frac{Y_f+t_k+t_c}{Y_m}$, and her relative weight in the household decision-making becomes: $\mu \left(v,\frac{Y_f+t_k+t_c}{Y_m}\right)$. The cash transfer directly enters the household budget constraint: $c^f+c^m+q=e^{\delta \left(v\right)}{w}_f{L}^f+n^f+w_m{L}^m+n^m+t_c$. The in-kind transfer directly enters the home production function:$Q=e^{\delta \left(v\right)}F\left(q+t_k,d^f\right)$.

15. Note that when status/instrumental motives are present and violence is distasteful, for violence to occur it must be the case that status/instrumental violence is already overpowering any bargaining power effects. Hence, it is unlikely (but not impossible) that a cash transfer to the woman will provide sufficient bargaining power to reduce violence.

16. Cudeville and Recoules (2015) introduce conformism to the bargaining process concerning a conjugal contract. Through the conformism of individuals, the conjugal social norm influences the marital behaviour of couples and the allocation of family resources. But the social norm itself results endogenously from the aggregation of couples’ marital agreements.

17. They find statistically significant increases in emotional abuse in households where physical violence did not increase.

18. Note however that as transfer size varies with household size, the paper cannot disentangle the effects of household characteristics and transfer size.

19. They also found a reduction in the reported frequency of violence by this group. They do not find an effect on emotional violence.

20. There was also evidence that the transfer + BCC recipients used more of the transfer for income-generating activities than participants who just received the transfer, and hence ended up with higher average wealth following the program. The higher wealth of this group may explain some of the difference between IPV rates in the two groups, if the transfer + BCC group had lower levels of household stress due to higher wealth.

21. Overall, the evidence of effects on cortisol is weak and measured using salivary cortisol, which is a noisy measure of stress levels. This study is not included in Table 2 as it does not report results for IPV independent of the attitudinal impacts.

22. A minority of recipients in Hidrobo et al. (2016) were men, but this was not randomised.

23. We omit Pettifor et al. (2016) from the meta-analysis as it studies a quite different form of cash transfer (to unmarried girls) and also reports the impact in terms of relative risk which isn’t readily comparable to the estimates in the other studies. We include two sets of estimates for Roy et al. (2019) – 1. For the transfer program alone; 2. For the transfer program with the behavioural intervention.

24. Excluding studies that report only long-term effects (Bobonis et al., 2015; Haushofer & Shapiro, 2018), studies without married or cohabitating partners (Pettifor et al., 2016), or studies that only looked at a subcategory of violence (Angelucci, 2008), we are left with 10 studies. Additionally, four studies bundled cash transfers with additional interventions like home visitation (Bobonis et al., 2013; Camacho & Rodríguez, 2016; Heath et al., 2020; Hidrobo et al., 2016).

25. Two papers looking at differences between sub-groups find results which are consistent with the existence of violence motivated by status inconsistency or instrumental motives (Angelucci, 2008; Hidrobo & Fernald, 2013). On the other hand, the lack of statistically significant differences in the impacts of transfers in the form of food and cash transfers in Hidrobo et al. (2016) suggests a limited role for status inconsistency or instrumental motives.

Additional information

Funding

An initial version of this paper was funded by the Department of Foreign Affairs and Trade, Government of Australia (grant number 73490). The views expressed in this paper are the authors’ alone and are not necessarily the views of the Australian Government.