Public Transfers and Spousal Violence

Preliminary – Please do not cite or quote




Public Transfers and Spousal Violence


Gustavo Bobonis*
Melissa González-Brenes†
Roberto Castro‡


December 2006


Abstract: Economic models of the family suggest that transfer programs in which funds are
targeted to women may increase the incidence of spousal abuse between partners. This study
uses data from a survey in Mexico to examine the impact of the Oportunidades conditional
cash transfer program on spousal abuse rates and the threat of violence. We find mixed
evidence regarding program impacts: although women in beneficiary households are 30
percent less likely to be victims of physical and sexual abuse than women in non-beneficiary
households, they are substantially more likely to receive violent threats with no subsequent
physical abuse. We present a model of asymmetric information in household bargaining to
document how increases in women’s income can lead to an increase in the amount of rents
that husbands are willing to extract, and in turn lead to a rise in their use of violent threats
with no subsequent physical abuse and a reduction in the actual use of spousal abuse –
predictions consistent with the empirical evidence.



A previous preliminary version of the paper was circulated as ‘Women’s Income, Female Status, and Spousal Violence:
Effects of the Mexican Oportunidades Program” (March 2006). We are grateful to Dwayne Benjamin, Morley
Gunderson, Deborah Reed, and Aloysius Siow, whose suggestions greatly improved the paper. We would also like to
thank Michael Baker, Gillian Hamilton, Heidi Shierholz, and seminar participants at Toronto and the PAA 2006
Conference for helpful comments, as well as Caridad Araujo, Tania Barham, Marta Rubio, Iliana Yaschine, and the staff
at Oportunidades for providing administrative data and their general support throughout. Research support from the
Institute of Business and Economics Research at UC Berkeley and the University of Toronto Connaught Fund is
gratefully acknowledged. We are responsible for any errors that may remain.

* Assistant Professor, Economics Department, University of Toronto. Address: 100 Saint George St., Room 4057,
Toronto, Ontario, M5S 3G3, Canada. Tel: 416-946-5299; E-mail: gustavo.bobonis@utoronto.ca.

† Assistant Professor, Department of Economics, University of Massachusetts-Amherst, Thompson Hall, Amherst, MA
01003, USA; Tel: 413-545-2012; E-mail: mgb@econs.umass.edu.

‡ Professor, Universidad Nacional Autónoma de México, Apartado postal 4-106, Cuernavaca, Morelos, 62431, México.
Tel: (52-777)-329-1853. E-mail: rcastro@servidor.unam.mx.

---

I. Introduction
Violence and the threat of violent actions are often used as mechanisms to redistribute resources in
society. In particular, spousal abuse has been proposed as a prominent example of economically
motivated violent behavior. In marriage, the quintessential example of an incomplete contract,
threats or actual acts of violence may be instruments at the disposition of individuals to bargain for
the control of resources or partner’s behavior in particular ways (Tauchen, Witte, and Long, 1991;
Bloch and Rao, 2002; Jacoby and Mansuri, 2006).
Moreover, marital violence is particularly prevalent across societies, and – as other forms of
violent behavior – it may be a very costly instrument of redistribution, making it an important public
policy issue. In the U.S., for instance, estimates of the prevalence of physical abuse during the 1980s
suggest rates of “severe” husband-to-wife violence of 30 per thousand per year (Blau, 1998).4 In our
context of study, rural Mexico, seven percent of currently married women reported being victims of
physical abuse from their male partners during the past year (INEGI, 2004). In Canada, evidence
from the early 1990s indicates that 29 percent of ever-married women and 50 percent of divorced
women have been the victims of spousal abuse. Recent attempts to measure the overall social cost
of domestic violence in Canada place it at around $4 billion, or 0.4 percent of GDP (Greaves,
Hankivsky, and Kingston-Riechers, 1995).5

An important policy question is: are there public policy mechanisms – for LDCs in particular
– which could be used to reduce the incidence of spousal abuse? A growing number of less
developed countries have introduced conditional cash transfer programs in which funds are
specifically targeted to women. This gender-based targeting of public transfers may have major
consequences for women’s welfare, since it is believed that resources in the hands of women
disproportionately improve their welfare (e.g., Thomas, 1990; Duflo, 2003). However, although one
of these programs’ objectives is the empowerment of women, increased prevalence of domestic
violence may be an unintended consequence, since unexpected changes in women’s income may
increase the incentives of male partners to use violence or threats of violence to extract rents from
their wives.
The objective of this paper is twofold. First, we provide evidence of the effect of the
Oportunidades conditional cash transfer program – a human development program initiated by the

4 Blau (1998) provides these estimates of “severe” physical violence.
5 As mentioned by Greaves, Hankivsky, and Kingston-Riechers (1995), putting a dollar value on violence against women
represents only one dimension of a very complex social problem; many aspects of violence against women, such as
emotional suffering, deterioration of the quality of or loss of life cannot be or are not easily quantified.

1

---

Mexican government in 1997 that provides cash transfers for marginalized households in rural areas
– on the prevalence of male-to-female spousal violence.6 To accomplish this, we use data from a
newly available nationally-representative survey, the National Survey on Relationships within the
Household (ENDIREH 2003), which includes detailed information on the prevalence and intensity
of male-to-female spousal abuse and threats of violence against women. We broadly define violence
from the survey measures as including physical, sexual, and emotional abuse, measures of spousal
violence that allow us to better characterize the real extent of spousal abuse among rural Mexican
households.
Constructing comparable groups of beneficiary and non-beneficiary households within each
village (to minimize the extent of omitted variable and selection biases), we find that women in
beneficiary households are 30 percent less likely to be victims of physical abuse than non-beneficiary
women, impacts that may come as a consequence of an increase in women’s empowerment within
the household. However, the distinction between the different types of violence is meaningful:
women in beneficiary households are as likely as non-beneficiary women to receive threats of violent
behavior and be victims of emotional abuse. Together, these results make a strong case for the
formulation of a theoretical framework that may help explain the divergence in the response of
husband’s use of threats of violence and the actual acts of violence as a result of changes in women’s
income.

As a second contribution of the paper, we present a theoretical framework adapting Bloch
and Rao (2002)’s asymmetric information model of domestic violence in household bargaining to
understand how male partners may use threats of and actual violence as instruments to extract rents
from their female counterpart’s share of the marital surplus. In this set-up, male partners signal their
dissatisfaction with marriage through the use of threats of violence and demand rent transfers from
wives if they have an incentive to do so. Acts of violence are then a response to the wives’ decision
to abide to the act of coercion. The model predicts that, if marginal increases in women’s income
lead to an increase in the amount of rents that husbands are willing to extract, then this will lead to
(i) an increase in the threat of use of spousal abuse with no subsequent physical abuse, and (ii) a
reduction in the actual use of spousal abuse. Finally, we test these empirical predictions of the model
using the variation in women’s income driven by the public transfer program. Essentially, we find
evidence consistent with the theoretical predictions: women in beneficiary households are

6 The transfer is paid to mothers contingent on certain requirements in terms of children’s school attendance and family-
level visits to health services.

2

---

substantially more likely to receive threats of use of spousal abuse with no subsequent physical abuse
than women in non-beneficiary households, and, as discussed above, less likely to be victims of
physical and sexually abusive behavior.
The findings outlined in this paper have important policy implications. Conditional cash
transfer programs are currently one of the main poverty-alleviation tools in Latin America and the
Caribbean, with programs providing transfers to mothers in Brazil, Colombia, Honduras, Jamaica,
and Nicaragua (Rawlings and Rubio, 2003; Maluccio and Flores, 2004). However, although women’s
empowerment is one of the programs’ objectives, domestic violence – with its potential negative
implications in terms of both women and child welfare – may be an unintended consequence. The
evidence presented provides a mixed view of their effectiveness in improving women’s
empowerment within the household, since the programs may increase the likelihood of violent
threats, which may in turn compromise women’s emotional health and other aspects of their
wellbeing.

This study contributes to a growing literature on how, due to the incomplete nature of
marital contracts, imperfect enforceability (e.g., limited commitment) and information asymmetries
among partners within the household may affect the efficiency of intra-household resource
allocation decisions.7 Ligon (2002), Basu (2006), and Lundberg and Pollak (2003) explore the
implications of limited commitment in dynamic bargaining models of the family and provide
conditions under which household allocation decisions diverge from full Pareto efficiency. Rasul
(2005) finds evidence consistent with limited commitment in partners’ fertility decisions among
partners in Malaysia, whereas Jacoby and Mansuri (2006) find evidence of parental household
strategies to limit the extent of ex-post marital discord among arranged marriages in Pakistan.
Finally, Ashraf (2006) and Bloch and Rao (2002) respectively address the role that information
asymmetries among partners in their incomes and levels of satisfaction affect household’s savings
decisions and spousal violent behaviors. The present study complements the literature by showing
how partners’ private information can help them gain undue influence in the behavior and welfare
of other household members.

7 A substantial empirical literature assesses the extent of full Pareto efficiency in intra-household resource allocation
decisions. Evidence from a number of studies in developed countries suggests that the intra-household allocation of
resources is constrained Pareto efficient (Browning et al. 1994; Browning and Chiappori, 1998; Chiappori, Fortin, and
Lacroix, 2002). However, Udry (1996), Dercon and Krishnan (2000), Duflo and Udry (2004), and Akresh (2005) find
evidence inconsistent with Pareto efficiency among rural household in Africa. These results are contested by Rangel and
Thomas (2005) using data for West African households and Bobonis (2006) among rural households in Mexico.

3

---

The paper is structured as follows. In Section II we briefly discuss the theoretical and
empirical literature on spousal abuse in both developed and developing countries. Section III
discusses the theoretical framework and the main predictions of the model. We present a concise
description of the Oportunidades program, its implementation, as well as the data used in the
analysis in Section IV. In Section V, we then describe our identification strategy and discuss how it
avoids the identification pitfalls. The main estimates are reported in Section VI, and Section VII
concludes.

II. Women’s Income and Spousal Abuse
There is a growing theoretical literature on the causes of domestic violence in economics and a rich
literature in other social sciences, particularly sociology and psychology. Spousal violence is
associated with both a source of gratification for the partner (husband) (e.g., a direct enjoyment of
the pain of another or release of frustration) and as an instrument for controlling the victim’s
resources or behavior (Tauchen et al. 1991; Bloch and Rao, 2002.). Existing models of spousal
violence include non-cooperative bargaining models in which female income, and more generally,
financial resources outside the marriage change the woman’s threat point and, all else equal, may
reduce the level of violence in equilibrium (Tauchen et al, 2001; Farmer and Thiefenthaler, 1997).
More recently, Bloch and Rao (2002) model spousal violence in India as a bargaining instrument
used to extract larger dowry payments from the bride’s family. However, the bargaining and
reallocation of resources takes place between the families of the bride and groom, rather than within
the couple itself.
Other theoretical work also addresses certain stylized facts of domestic violence, including
the fact that battered women are not unlikely to return to an abusive relationship even after seeking
help.8 In a model by Farmer and Thiefenthaler (1996), battered women use shelters and other
support services to signal to the abuser their ability to leave the relationship, which changes their
threat-point and may reduce their toleration for physical abuse. Finally, Pollak (2003) addresses the
intergenerational transmission of violent behavior within the household. In the model, individuals
raised in violent homes are more likely to marry partners who were also raised in violent homes.
Thus, assortative matching may increases the equilibrium level of violence.
At risk of oversimplifying, sociological models link violence to gender inequality. The causes
of gender violence lie in the way society is organized: for instance, unequal economic opportunities

8 See Bowlus and Seitz (2005) for an exception to this common belief in the Canadian context.

4

---

available to women and men, the availability of institutional resources for women who are victims of
spousal violence, and the degree of protection offered by the legal system all affect the prevalence of
violence against women.9 Psychological models also incorporate individual characteristics as
determinants of violence. The literature characterizes violence as an expression of the batterer’s
desire for control over the victim, and links violence to batterer’s low self-esteem, pathological
jealousy, and severe stress. There is also qualitative evidence that abuse is often accompanied by a
curtailment of the victim’s economic and social independence.10
Existing (non-experimental) empirical evidence on the relationship between female and
income and violence is mixed. In the U.S., Tauchen et al. (1991) examine this relationship using a
sample of 125 women referred from shelters and other advocates for battered women, and find the
expected negative correlation between violence and female income for a subset of low and middle
income couples in their sample. However, these results are based on a small non-random sample (of
battered women), and therefore are not comparable to studies based on representative samples of
women. Stevenson and Wolfers (2006) examine how a particular change in women’s opportunities
outside current marriages - unilateral divorce legislation in the United States - changed patterns of
family violence and whether the option reduced female suicide and spousal homicide. They find
evidence consistent with unilateral divorce laws substantially reducing all of the above outcomes. In
the context of less developed countries, Panda and Agarwal (2005) find that women who own
property are less likely to be victims of spousal violence in India, but González-Brenes (2005) does
not find a relationship between female income shares and the probability of violence for women in
several East African countries. In the next section, we discuss a model of asymmetric information in
household bargaining to document how increases in women’s income can lead to a rise in partners’
use of violent threats and reductions in the actual use of spousal abuse – predictions that may
address the potential paradoxes presented in the paper and those existing in the empirical literature.

III. Husband’s Private Information, Violent Threats and Spousal Abuse 11
The following theoretical framework adapts Bloch and Rao (2002)’s signaling model of domestic
violence to understand how male partners may use threats of and actual violent behavior as
instruments to extract rents from their female counterpart’s share of the marital surplus. This set-up

9 For details on the sociological literature, see the review in Castro (2004).
10 See Walker (1984).
11 This section draws heavily on Bloch and Rao (2002). Essentially, we provide a different interpretation to their model
of terror as a bargaining instrument across families.

5

---

will allow us to uncover the mechanisms through which increases in wives’ income may result in
greater threats of spousal abuse from their partners but in a lower incidence of actual acts of abuse.
We consider a static non-cooperative bargaining model with asymmetric information, where
the threat of violence is interpreted as a signal sent by the husband to the wife about his level of
satisfaction with the marriage. This asymmetry allows the husband to demand a resource transfer
from his partner. Following the latter’s decision whether to abide to the former’s demands or not,
the model specifies the conditions under which actual acts of physical violence will occur. Although
concerns may be raised as to whether repeated (complete information) non-cooperative bargaining
models could provide similar predictions regarding violent behavior, it is a priori possible that these
are likely to sustain efficient outcomes, ruling out the exercise of threats and violence in equilibrium.
More specifically, our application of the Bloch and Rao (2002) model results in an
equilibrium in which, after the husband’s level of satisfaction has been revealed, (i) satisfied
husbands would prefer to remain in the current union independently of receiving transfers from
their partners and do not threaten with or use violence against their partners; (ii) unsatisfied
husbands would prefer to use violence and reach a Pareto inefficient intra-household resource
allocation unless they receive a transfer of the marital surplus from their wives; they use threats of
violence to signal their dissatisfaction, and may obtain transfers from their partner. We discuss the
set-up and main predictions of the model in the following paragraphs.
We will analyze the distribution of resources in a two breadwinner household where h and w
respectively denote the two agents: husband and wife. The marriage forms based on the potential
gains to marriage – which may arise from specialization in home and market production, the joint
production or provision of household public goods – and the feasible division of the marital surplus.
The union leads to indirect utility levels for each partner given by U = u (I , x , x , ?) and
h
h h
h
w
U = u (I , x , x ), where I , I represent husband’s and wife’s incomes, x , x , denote vectors of
w
h w
h
w
h
w
h
w
human-capital characteristics of each partner, and ? is the husband’s private level of satisfaction with
the marriage. Assume that the indirect utility functions are strictly increasing in all their arguments
and strictly concave in income.
We assume that the negotiation game has the following structure:

Structure of the game:
Stage 1: The quality of the match is revealed, both public (z) and private (?) components;

6

---

Stage 2: The husband chooses to threaten the wife with the use of physical violence and demands
a transfer;
Stage 3: The wife responds to the husband’s demand by accepting or rejecting to provide the
transfer amount demanded;
Stage 4: The husband chooses whether to use physical violence or not and, if so, the
partnership’s equilibrium allocation choice is Pareto inefficient (e.g., “separate spheres”
equilibrium).

We will now discuss in detail the assumptions made in the structure of the game. In the first stage,
the quality of the match is revealed. This includes a public component, z, observable by both
partners, and a private component, ?, which is only observed by the husband. We assume that the
private component is a dichotomous variable, with value 1 for satisfied husbands and 0 for
dissatisfied ones. The prior probability that the husband is dissatisfied, Pr(? = 0), is a function of the
observable characteristics of the marriage, p(z), with p?(z) < 0.
In the following stage, the husband chooses whether to threaten the wife with physically
abusing her. Both partners suffer a utility loss measured by C (?) and C if the threat of violence is
h
w
made. This relies on the idea that threats of physical violence are actual incidences of emotional
violence, and these may be prejudicial to both men’s and women’s mental health and emotional
wellbeing. Also in this stage, we assume that the husband has all the bargaining power, and makes a
take-it-or-leave-it demand of a transfer t to his wife. This assumption, in which bargaining power
radically shifts in favor of the man once the woman commits herself to marriage, may represent the
true extent of women’s bargaining power in many traditional societies, in which women’s formal
legal rights are often weak and divorce is highly stigmatized (Jacoby and Mansuri 2006; Bobonis
2006). Following the possible threat and demand, the wife responds by accepting or rejecting the
offer. In the current version of the model, we assume that divorce is too costly for the wife and she
will prefer to pay a transfer than divorce, but will in the future endogenize the wife’s divorce option
at this point.
In the final stage of the game, the husband chooses to actually use physical violence to
punish his wife’s potential deviant behavior, which entails the destruction of a share of the marital
surplus, and the household reaches a non-cooperative “separate spheres” equilibrium that in may
additionally reduce the marital surplus. If this were the case, the partners enjoy more limited joint

7

---

production possibilities of the marriage, and obtain discounted utilities denoted V = v (I , x , x ) and
h
h h
h
w
V = v (I , x , x ), both strictly increasing in its arguments and strictly concave in income.
w
w w
h
w
The following assumptions - analogous to Bloch and Rao (2002)’s – ensure a uniquely
determined separating perfect Bayesian equilibrium of the game (which satisfies the Cho-Kreps
intuitive criterion), where, as mentioned above, satisfied husbands do not make violent threats or
use violence, whereas unsatisfied husbands have incentives to use violent threats to extract a transfer
from their partners.

Assumption 1: For any level of income (I ) and socio-economic characteristics of the partners (x ,
h
h
x ), u (I , x , x , ?=1) > v (I , x , x ) and u (I , x , x , ?=0) < v (I , x , x ).
w
h h
h
w
h h
h
w
h h
h
w
h h
h
w

Assumption 2: Husband’s cost of making violent threats. C (1) = ?; C (0) = ? is a random variable
h
h
with cumulative distribution function F on [0, ?).
?

Assumption 3: A wife will strictly prefer to suffer violent threats and provide transfers than to
reject the partner’s demand, suffer physical abuse, and reach a Pareto-inefficient intra-household
resource allocation ourcome: u (Y - t, x , x ) - C ? v (Y , x ).12
w
w
h
w
w
w
w
w

We briefly discuss the conditions which characterize the equilibrium, and which elucidate how
increases in women’s income may affect their partner’s violent behavior. First, the wife’s decision to
accept or reject the demand will depend on the expected cost – based on her posterior beliefs
regarding the probability that the husband is dissatisfied, ? (given the threat of violence) – of paying
a transfer to a husband who is satisfied, relative to the prospects of suffering physical abuse she
faces from a dissatisfied husband if she rejects the demand. Formally, the wife will accept to pay any
transfer payment demand such that:

(1)
u (Y - t, x , x ) ? ? v (Y , x ) + (1-?) u (Y , x , x )
w
w
h
w
w
w
w
w
w
h
w

The maximal amount that the wife is willing to pay to her husband, tw(?), is the solution to the
following condition:

(2)
? v (Y , x ) + (1-?) u (Y ,x , x ) = u (Y - tw(?), x , x )
w
w
w
w
w
h
w
w
w
h
w


12 In future work, we will endogenize the wife’s divorce decision by allowing another option for the wife: leave the
relationship and choose the level of transfer payments in this case (possibly zero). This will probably be the case as long
as uw(Yw - t, xh, xw) - Cw < vw(Yw, xw).

8

---

Next we consider the husband’s incentive to make violent threats and the determinants of the
amount of transfer demanded. In equilibrium, satisfied husbands will never make violent threats,
since the cost of threatening the wife with physical abuse is infinite (by assumption). Among
dissatisfied ones, individuals whose cost of threats of violence is low enough will demand the
highest possible transfer from their spouses, tw(1), and alternatively those whose costs of violent
threats is too high will not make transfer demands and will behave so as to achieve the inefficient
intra-household household allocation equilibrium. Specifically, there will be a threshold value of the
cost of making violent threats (?*) for which a dissatisfied husband is indifferent between making
the threat and obtaining the transfer tw(1) and using violence.. The value of ?* is determined by:

(3)
?* = u (Y + tw(1), x , x , ?=0) – v (Y , x ).
h
h
h
w
h
h
h

Therefore, the probability of violence given that the husband is dissatisfied is Pr[? < ?*] = F (?*),
?
and the unconditional probability that the husband will be violent is:

(4)
B(Y , Y , x , x , z) = p(z) F [u (Y + tw(1), x , x , ?=0) – v (Y , x )]
h
w
h
w
?
h
h
h
w
h
h
h

Conditions (2) and (4) allow us to assess how a change in women’s income may affect the incidence
of violent threats and actual physical abuse in the relationship. A marginal increase in the wife’s
income will lead to a change in the maximal transfer amount demanded by her partner that will
make her indifferent between accepting or rejecting the latter – this change could be positive or
negative. By implicit differentiation of equation (2),

? ?u
? ?v
?u
??
? w (Y , x , x ) + ?
Y x x
Y x x
w
h
w
w
?
? w ( , , ) ? w ( , , )
w
h
w
w
h
w
?
??
(5)
?t
? ?I
I
I

w
? ?
?
w
w
??
=1?
?
?
Y
?u
?
w
?
w (Y ? t, x , x )
w
h
w
?
?
?
?Iw
?
?

Equation (5) suggests that the change in transfer amount demanded ( w
t
?
Y
? ) is increasing in the
w
change in wife’s income if the utility gain from accepting the transfer is sufficiently greater than the
“utility gain from achieving the Pareto efficient intra-household allocation” relative to her utility
from the inefficient “separate spheres” allocation, or:

(6)
1 ? u
?
u
?
?
v
?
u
?
w (Y ? t, x , x )
w
?
(Y , x , x )
w
>
(Y , x , x )
w
?
(Y , x , x )
?

w
h
w
w
h
w ?
w
h
w
w
h
w
? I
?
I
?
I
?
I
?
? w
w
?
w
w


9

---