Measuring Women's Empowerment: lessons to better understand domestic violence

Full text

Measuring Women’s Empowerment: lessons to better
understand domestic violence
Diana Lopez-Avila
To cite this version:
Diana Lopez-Avila. Measuring Women’s Empowerment: lessons to better understand domestic
violence. PSE Working Papers n2016-04. 2016. <halshs-01294565>
HAL Id: halshs-01294565
https://halshs.archives-ouvertes.fr/halshs-01294565
Submitted on 29 Mar 2016
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci-
entific research documents, whether they are pub-
lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destin´ee au d´epˆot et `a la diffusion de documents
scientifiques de niveau recherche, publi´es ou non,
´emanant des ´etablissements d’enseignement et de
recherche fran¸cais ou ´etrangers, des laboratoires
publics ou priv´es.

WORKING PAPER N° 2015 – 04
Measuring Women’s Empowerment:
lessons to better understand domestic violence
Diana Lopez-Avila
JEL Codes: D13, I15, J12, J16, O12
Keywords: Gender, Domestic Violence, Household bargaining models, Social
Capital
PARIS-JOURDAN SCIENCES ECONOMIQUES
48, BD JOURDAN – E.N.S. – 75014 PARIS
TÉL. : 33(0) 1 43 13 63 00 – FAX : 33 (0) 1 43 13 63 10
www.pse.ens.fr
CENTRE NATIONAL DE LA RECHERCHE SCIENTIFIQUE – ECOLE DES HAUTES ETUDES EN SCIENCES SOCIALES
ÉCOLE DES PONTS PARISTECH – ECOLE NORMALE SUPÉRIEURE – INSTITUT NATIONAL DE LA RECHERCHE AGRONOMIQU

Measuring Women’s Empowerment: lessons to better understand
domestic violence∗
Diana Lopez-Avila†1
1Paris School of Economics
This Version: March 28, 2016
Abstract
This paper aims at shedding light on the relationship between women’s empower-
ment and domestic violence. For this, we explore different ways to measure women’s
empowerment and domestic violence, and analyze whether the relation depends on the
definitions used. We take advantage of a rich data set collected in rural Colombia,
including several measures of self-esteem, disagreement towards domestic violence, par-
ticipation in household decisions and social capital; and analyze the relationship with
both aggressive and controlling ways of domestic violence. The results indicate that the
different measures of women’s empowerment help explain much better the aggressive
ways of domestic violence than the controlling ones. Our results show a positive cor-
relation between women’s empowerment and domestic violence. This goes in line with
the theories that argue that men use violence as a way to leverage their power within
the household. Among the different latent measures of women’s empowerment we used,
we found that social capital and self-esteem are significantly correlated with aggressive
domestic violence. We do not find that more common proxies, such as women’s partic-
ipation in household decisions, are significantly correlated to domestic violence.
Keywords: Gender, Domestic Violence, Household bargaining models, Social Capital
JEL classification: D13, I15, I31, J12, J16, O12
∗I’m particularly grateful to Karen Macours for her thorough guidance in this work. I also thank Andres Moya at Universidad
de los Andes in Colombia, who was in charge of the project in which the data for this research was collected. I also thank Sylvie
Lambert and the participants to the CFDS seminar at PSE for their useful comments. I also thank the external members of my
thesis committee, Orazio Attanasio and Raquel Bernal for their useful comments. Any errors are mine.
†diana.lopez@psemail.eu
1

Introduction
Violence against women constitutes an obstacle for social and economic development. Domes-
tic violence is associated with health and psychological problems for women [Ellsberg et al.,
2008], [WB, 2014]; and adverse consequences for children later in life [Unicef, 2014]. Boys
who witnessed their mother being beaten are at a higher risk of becoming violent part-
ners [Heisse, 2011]; and moreover, families that experience spousal violence are more likely
to experience violence against children [Herrenkohl et al., 2008], [WB, 2014]. Being vic-
tim of domestic violence can also have important economic consequences. Severely abused
women earned 61% less in Chile and 43% less in Nicaragua, as compared to non-abused
women [Morrison and Orlando, 1999]. Estimations on the impact of domestic violence on
GDP in Colombia, due to women’s lower earnings, suggest a cost of approximately 3.2%
[Ribero and Sanchez, 2004].
Despite the fact that domestic violence is a serious problem that constrains women’s
economic and social development, understanding the correlates of domestic violence is not
straightforward. Women’s bargaining power and domestic violence are closely intertwined,
and defining women’s bargaining power can be complex. This paper contributes to the
literature by providing descriptive evidence on the relationship between women’s empow-
erment and domestic violence. As a starting point, we compute aggregate measures of
women’s empowerment and domestic violence, taking into account the different dimensions
we captured in our survey. We show a positive relationship between women’s empowerment
and domestic violence, which can seem a bit puzzling. This goes contrary to the common
idea that more empowered women are more able to stand for their rights and hence, less
exposed to domestic violence. However, economic and sociological theories provide argu-
ments on why the relationship between domestic violence and women’s empowerment can go
in both directions, [Farmer and Tiefenthaler, 1997], [Tauchen et al., 1991] [Bloch and Rao,
2002], [Eswaran and Malhotra, 2011], [Heisse, 2012].
To better understand and dig deeper into this relationship, we implement a factor anal-
ysis to disentangle the different latent dimensions of empowerment and domestic violence.
We explore dimensions of women’s empowerment such as self-esteem, disagreement towards
domestic violence, willingness to divorce, participation in household decisions, social capital,
income and education. We also differentiate the aggressive ways of domestic violence from
the controlling behavior from the partner. Drawing on the different theories on domestic
violence, we include different socio-demographic characteristics that the theory predict to be
correlated to domestic violence, such as woman’s and partner’s income, age and education;
participation in the labor market, type of union, and experience of domestic violence as a
child, among others. Evidence found in this paper helps to better characterize the relation
between women’s bargaining power and exposure to domestic violence, by suggesting which
can be the proxies that better describe women’s autonomy. We interpret the empirical find-
ings within the theoretical models that have analyzed domestic violence.
Measuring women’s bargaining power is not straightforward. Women’s empowerment
has multiple dimensions [Alkire, 2007]. Usually, concepts such as access, ownership, entitle-
1

ment and control are used interchangeably [Kabeer, 1999]. In a recent work, the Food Pol-
icy Research Institute (IFPRI) and the Oxford Poverty and Human Development Initiative
(OPHI), tried to encompass the different dimensions and spheres of women’s empowerment
in agriculture [Alkire et al., 2014], [Alkire et al., 2013]. They identify five main domains of
empowerment: decisions about agricultural production, access to and decision making power
over productive resources, control over use of income, leadership in the community, and time
allocation. Although this index is a step forward to better measure women’s empowerment,
it has some limitations. Women who do not actively participate in agriculture may appear
less empowered, as they take decisions in non-agricultural issues. On the other hand, in
households where there are no males, women may appear more empowered, as they do not
have to whom to bargain for household decisions.
A common way in which researchers have approached women’s empowerment is through
their participation in household decisions. Usually, surveys contain a set of questions asking
about who is in charge of the income use, the type of food consumed, children’s schooling or
the use of contraceptive methods, among others. However, deciding which type of food to buy
may not have the same consequences as deciding whether or not to send children to school,
or whether or not to use contraceptive methods. Besides, it is not clear whether sole decision
making implies more empowerment than jointly decision with the partner. Usually, joint deci-
sion is the most common answer, leaving limited variation across answers [Almas et al., 2015].
This limited variation may help explaining why social programs, such as CCTs, do not seem
to be drivers of women’s empowerment. For Progresa in Mexico, results show a reduction in
the likelihood of sole decision-making by husbands with regard to medical treatment, school
attendance, and child clothing, due to the program; but no strong impact on the likelihood
that women solely, or jointly with their partners, decide [Adato et al., 2000]. For Familias
en Accion in Colombia, results suggest that there is no change in women’s decision power,
[Camacho and Rodriguez, 2012]. Authors analyze household’s decisions on children’s health,
children’s school attendance, clothing expenditure for children, food expenditure and the use
of additional income in the household. For none of the outcomes a significant impact is found.
In a recent work, [Almas et al., 2015] use a novel way to measure women’s empowerment
within the household. In a lab setting in Macedonia, they measure women’s willingness to
pay to receive cash transfers instead of their partners. Besides, authors take advantage of the
random assignment of a CCT to the household heads or mothers, at the municipality level.
Authors find that women are willing to sacrifice some household income in order to receive
the cash transfer. However, this willingness is lower in municipalities where women have
already been empowered by the CCT (eg. municipalities where women were assigned to re-
ceive the CCT). Following a similar line, [Peterman et al., 2015] analyze survey experiments
undertaken in cash and food transfer programs in Ecuador, Yemen, and Uganda. Authors
find large variations in how women are ranked in terms of decision making depending on
how questions are framed and indicators are constructed. Besides, authors find that deci-
sion making indicators are not consistently associated with other proxy measures of women’s
empowerment, such as women’s education. The only factor that matters across countries is
women’s age, which is positively correlated to the decision making power.
2

Given the difficulty to accurately measure women’s empowerment and domestic violence,
the empirical evidence on the association between women status within the household and
domestic violence is mixed, as well as the evidence on the impact of social programs on do-
mestic violence. Women with partners who have education secondary or higher are less likely
to experience domestic violence in countries such as Egypt and India, but more likely in coun-
tries like Peru. Receiving cash (while not working) is negatively associated with experiencing
domestic violence in Egypt, but positively associated in India and Peru, [Kishor and Johnson,
2004]. Female dominated decision making is associated to higher levels of domestic violence,
in countries such as Peru and Haiti, [Flake, 2005], [Gage, 2005]. Regarding financial inclusion
and domestic violence, results from the IMAGE study in South Africa and a group-based
credit program in Bangladesh, show that treated women are less likely to experience domestic
violence, [Kim et al., 2007], [Schuler et al., 1996].
In the case of the Oportunidades program in Mexico, results show that beneficiary women
are less likely to be victims of physical violence, but more likely to receive violent threats
with no associated physical violence [Bobonis et al., 2013]. Another study shows that there
is a decrease in violence towards women receiving small transfers; while there is an increase
towards those receiving large transfers, particularly with partners with traditional views of
gender roles, [Angelucci, 2008]. Results from a unconditional cash transfer in rural Ecuador
suggest that women’s education matters for the direction of the impact. There is a significant
decrease in psychological violence for women with education greater than primary. However,
there is an increase in emotional violence for women with primary education or less, but
higher or equal than that of their partners [Hidrobo and Fernald, 2013].
Other socio-demographic characteristics have also shown to be associated with domes-
tic violence. Young and less educated women seem to be at an increased risk of domestic
violence [Heisse, 2012]. Living with a partner without formal marriage also appears to in-
crease women’s risk of domestic violence [Kishor and Johnson, 2004], [Flake, 2005]. Accord-
ing to data from the Demographic and Health Survey of 2010 in Colombia, women in a civil
union report higher rates of emotional and physical violence, [Ojeda et al., 2011]. The larger
the number of children in the household, the higher the women’s risk of domestic violence,
[Heisse, 2012]. Exposure to violence during childhood also appears to increase the likeli-
hood of experiencing domestic violence as an adult, as well as partner’s alcohol consumption,
[Kishor and Johnson, 2004]. Couples where both partners work appear to be at a (slightly)
higher risk, as compared to couples where only the man works, [Heisse, 2012].
Previous empirical evidence shows that the relation between women’s status and domestic
violence can go in different directions. The economic, sociological and psychological theories
have also shown that the relation between women’s empowerment and domestic violence can
be positive or negative. Sociological theories point out that women’s and partners’ resources
are important at determining domestic violence, suggesting that women with higher status in
the household are at a higher risk of abuse. The resource theory states that men with fewer
resources would be more likely to use violence. Men facing an economic crisis or unemploy-
ment, may be more vulnerable to perpetrate abuse [Goode, 1971]. Related to this theory is
the relative resource theory, which states that women with higher status than their partners
3

will be at a higher risk of abuse. Men use force or psychological manipulation to reaffirm
their dominant status [Macmillan and Gartner, 1999]. The status inconsistency theory states
that wife beating may be high in settings where men’s traditional power in the family has
declined, while women’s power has increased [Levinson, 1989]. In a similar line, the gender
role stress theory argues that when economic or social conditions make men feel powerless,
they exert power where they still can: the family [Jewkes, 2002]. From the psychological
side, exposure to violence and poor parenting in childhood can influence the likelihood of
future violence. This influence can be through modeling or through the negative impacts on
child development [Dutton, 1995].
Family as an economic unit has been analyzed for several decades. The economic the-
ories that have modeled domestic violence have approached it through a non-cooperative
setting. One set of theories predict that an increase in women’s reservation utility leads to
a decrease in domestic violence. In a setting when men can buy domestic violence through
the monetary transfer they make to women, an increase in women’s wages, hours worked or
non-wage income, is going to decrease the value of the transfer; hence, the level of domestic
violence [Farmer and Tiefenthaler, 1997]. Similarly, [Tauchen et al., 1991] argues that men
use violence towards women for a behavior of which they do not approve. He predicts that
an increase in the woman’s income is going to lead to a decrease in violence, while a rise in
the husband’s income leads to an increase. Empirical evidence shows that this is true for
all households, except the highest income ones. These economic approaches have also the
support from the sociological theories. The fewer resources a woman has, the less power she
has, and the less likely she is to leave an abusive relationship [Heisse, 2012].
Another set of theories allow domestic violence to affect women’s bargaining power.
Within this framework, the extent of domestic violence faced by women is not necessarily de-
clining in their reservation utilities, or necessarily increasing in their spouses [Bloch and Rao,
2002], [Eswaran and Malhotra, 2011]. Greater domestic violence may also be a rational male
response to the greater autonomy of women. These approaches go in line with the evolution-
ary theory, which states that domestic violence originate from paternity uncertainty [Heisse,
2012]. Men could think that more independent women will have more sexual contact with
other men. In this setting, spousal violence is intended to increase the abuser bargaining
power. [Bloch and Rao, 2002] propose a model of asymmetric information, in which the
husband uses violence to signal to his in-laws the degree of satisfaction with the marriage.
Women from wealthier families are more likely to be beaten, as violence is used as a weapon
for extortion.
This paper aims to shed light on the relation between domestic violence and empower-
ment. We take advantage of a data set collected to evaluate a rural development program,
Oportunidades Rurales (OR), and the possible synergies it could have with a Conditional
Cash Transfer (CCT), Familias en Accion.Oportunidades Rurales is a rural development
program that provides technical and financial assistance to small rural entrepreneurs. Women
who participate in these organizations have to attend meetings regularly, having access to
more information and higher contact with other women. We implement a factor analysis to
get the different measures of women’s empowerment and domestic violence. The women’s
4

empowerment factor analysis suggests four factors to model women’s empowerment. One
factor corresponds to measures referring to self-esteem and disagreement towards domestic
violence. A second factor corresponds to intra-household decision making. A third factor,
to the role of women in the productive organization receiving Oportunidades Rurales; and
a fourth factor, to the women’s social network. The domestic violence factor suggests two
factors, one mainly loading on aggressive behavior, and a second one on controlling behavior
from the partner.
Our results suggest that the different measures of women’s empowerment help explaining
much better the aggressive ways of domestic violence, rather than the controlling ones. More-
over, social capital, together with women’s self-esteem, appears to better capture women’s
autonomy. These two dimensions of women’s empowerment are positively correlated to do-
mestic violence. On the contrary, we do not find women’s participation in household decisions
to be correlated to domestic violence. Among the socio-demographic characteristics, living
in a rural area or having a larger number of children in the household, is positively corre-
lated with experiencing aggressive ways of domestic violence; while a larger household size is
negatively correlated. Results found in this paper go in line with the economic models that
predict an increase in domestic violence due to an increase in women’s reservation utility. In
these cases, partners use violence as a way to leverage bargaining power within the household.
The rest of this paper is organized as follows. Section 1 describes the data and the frame-
work of the interventions for which it was collected. Section 2 provides some descriptive
evidence motivating the analysis. Section 3 provides the results together with the robustness
checks; and Section 4 concludes.
1 Data
In 2013, as part of a initiative of The International Fund for Agricultural Development
(IFAD), the Universidad de los Andes in Colombia collected data to evaluate the impact of
Oportunidades Rurales and the possible complementarities it could have with Familias en
Accion.Oportunidades Rurales (OR) is a rural development program financed by the Interna-
tional Fund for Agricultural Development (IFAD) and launched in 2007. Oportunidades Ru-
rales aims at increasing productive, human and financial assets of small rural entrepreneurs,
through technical and financial assistance. The program has a demand-driven approach, in
which rural productive organizations and households need to identify their specific needs and
use of funds. Between 2007 and 2013, 1,817 organizations and 47,018 households benefited
from the program. Organizations belong to 714 municipalities in 25 (out of 32) departments
of the country1, [Econometria, 2014]. The selected organizations received joint financing for
technical assistance in productive (inputs or physical capital), commercial (packing or image
of products), or administrative aspects. They also participated in commercial fairs at the
local and national level. OR targets productive organizations that gather farmers belonging
to the poorest population of the country. Priority was also given to organizations of young
1Departments such as Huila concentrated a large number of beneficiary organizations.
5

people and women. In terms of geographic location, the program was prioritized to rural
areas with high poverty rates, presence of violence and displaced population; but also with
high presence of farmers’ organizations.
Familias en Accion is nowadays the largest social program in the country2. It has been
implemented since 2002 and has the traditional components of a Conditional Cash Transfer
(CCT), nutrition, health and education. The nutrition component consists of a monetary
supplement given to all beneficiary families with children under 7 y.o. The health component
consists of vaccination and growth and development checks for children, and courses of nutri-
tion, hygiene and contraception for mothers3. The education component is given to families
with children 7 to 17 y.o., and the amount of the subsidy changes between primary and sec-
ondary. The transfer for a family that received the education, for primary and secondary, and
the nutritional subsidy, will represent approximately 35% of its income4. Families belonging
to Sisben 1 and with children 0-18 y.o are eligible for the program. Familias en Accion has
shown to be successful in improving children’s nutrition, health and education. The program
has increased protein consumption, school enrolment and health outcomes, such as height
for age and weight for age, [Attanasio and Mesnard, 2006; Attanasio et al., 2012].
Data were collected for 59 organizations between January and June 20145. The survey
was designed following the Colombian Longitudinal Survey of Wealth, Income, Labor and
Land (ELCA) carried out by Universidad de los Andes since 2007. The survey included
modules on consumption, production, assets ownership, labor, education, time uses and food
security, among others. Two additional sections were included, one referring to expectations
and another to gender.
The gender section was specifically designed for this study. It was designed building on
the domestic violence section of the Demographic and Health Survey (DHS), and the gender
section of a survey to evaluate the impact of a CCT in Malawi6. The gender module includes
three sections. One in which women were asked about their self-esteem, and their agree-
ment/disagreement towards domestic violence. A second one on household decision making
on income uses, expenditures in clothing and education, and contraceptive methods. The last
section corresponds to the domestic violence one, in which women were asked about occur-
rence of different events of violence. The events of domestic violence include emotional and
physical violence. Emotional violence includes situations such as: has your partner accused
you of being unfaithful, has humiliated you, has threatened to leave you for another woman,
has threatened to hit you, does your partner not allow you to meet your family and friends,
2The program covers all municipalities and 2.6 millions of families. Taken on July 27th of 2014 from
http://www.dps.gov.co/Ingreso_Social/FamiliasenAccion.aspx
3The participation in the health checks and courses is compulsory to receive the nutritional subsidy
4The conditionality and amounts of the transfers changed in 2013, when the program was redesigned,
being now known as Mas Familias en Accion
5Sampling was done using year of entering into the program, productive activity and geographical location.
Also, the score the organizations got at submitting their proposals was taken into account. Members of the
organization were randomly selected within each organization.
6I thank Berk Ozler, at the World Bank, who shared with me the household survey for the Schooling,
Income and Health Risk project in Malawi
6

has insisted in knowing where you were all the time. Physical violence includes events such
as: has your partner pushed you or hit you, or has forced you to do things you did not want
in the intimacy. This section was asked under the condition of complete privacy and all
enumerators were women. The section also includes questions on willingness to get divorced
and on the reasons women consider domestic violence started or increased.
The quantitative data collection was complemented with qualitative work to get some in-
sights on how women’s participation in the organization may interact with domestic violence.
We ran focus groups with male and female members of the associations. The gender module
developed during the focus groups included a discussion on gender roles and women’s par-
ticipation in the organization. We also did some semi-structured interviews to rural women
to approach how common it was to be victim of domestic violence, how open they were to
talk about it, what kind of events of domestic violence they experienced and, what were
the reasons they expressed for the events of violence to occur. During this qualitative work,
women expressed that episodes of domestic violence were highly associated with partner’s
alcohol consumption. They also expressed that before the episodes of physical violence took
place, their partners showed a possessive behavior. Partners appear to be jealous of women
going out often and meeting other women7.
The data set contains a rich module, asked to women in a confidential setting, specifically
designed to explore the relation between women’s empowerment and domestic violence. We
approach women’s empowerment in different ways, self-esteem and acceptance of domestic
violence, participation in household decisions and, social capital; as well as the traditional
measures, such as education and income. Following, [Kabeer, 1999], women’s empowerment
can be seen in three big inter-related dimensions: resources, agency and achievements. We
consider that our measures will shed light on these different dimensions of women’s empower-
ment. We also collected information on events of aggressive domestic violence and controlling
behavior. Moreover, we collected information on women’s desire to get divorced and the rea-
sons associated; as well as on the reasons they consider domestic violence started or increased.
Information on other characteristics that may affect domestic violence is also available, such
as type of union, labor force participation and having witnessed domestic violence as a child.
This comprehensive set of information gives us the opportunity to shed light on the relation
between women’s empowerment and domestic violence, and the sensitivity of the relationship
to different measurement choices.
For self-esteem and acceptance of domestic violence, we use a set of variables asking about
how proud women were of themselves and how capable they saw themselves of doing different
things. We also ask about whether women agree or disagree on traditional gender roles, such
as the fact women should always obey their partners, or that the use of domestic violence
could be accepted under some circumstances. For household-decision making, we use a set
of questions asking about the main decision maker on household consumption and income
7A 40-year old woman from Espinal-Tolima expressed the following: “I think most of the cases of domestic
violence happen because husbands drink (alcohol). They always accuse women of being unfaithful... A lot
of husbands do not like when women participate in meetings with other women, such as the ones done by
Familias en Accion. They say that women go there to gossip and to meet who knows who”
7

use. For social capital, we characterize the role of women in the productive organization
that received Oportunidades Rurales, as well as their social networks. In the case of domes-
tic violence, we asked if women have ever experienced different events of domestic violence.
These events included aggressive ways of violence, such as hitting or threat to hit or abandon
for another woman; and controlling ways, such as accusation of being unfaithful or limiting
contact with families and friends.
For those who were members of the organization, a social capital module was asked. This
module included questions on the actual participation in the productive organization, as well
as on the participation in other type of organizations. Members were asked about whether
or not they were still part of the organization, the time they have been participating, their
role in the organization, the frequency in which they attend the meetings, whether or not
they participate in the decision process and the reason that motivated them to participate in
the organization. They were also asked in which other organizations they participate, such
as Communal Action Boards or Parents Associations; and whether or not they were leaders
of these organizations.
The final sample for which data was collected, corresponds to 729 households and 59
organizations located in the Departments of Boyac á, Caldas, Cundinamarca, Huila, Risar-
alda and Tolima. Organizations included in the sample account for, approximately, 36% of
the total of programs financed my OR in the period of analysis. The productive activities
done by the organizations correspond mainly to dairy, fruits and minor species. Among the
59 organizations, 16 received OR in 2008-2009 and 43 in 2012-2013. Some of the surveys
took place at the headquarters of the organization, which implied that it was not possible to
complete the gender module due to lack of privacy. Because of this, among the 729 house-
holds interviewed, only 283 answered to the gender section8. Among the 283 women, 198 are
members of the organization, while 85 are spouses of members of the organization.
2 Descriptive Analysis
Given that only a bit more of one third of the sample answered to the gender section, some
selection concerns may arise, though the reason to not answer to the gender section was not
because the woman declined to answer the section. Table A.1 shows some basic statistics for
the whole sample, and Table A.2 shows the differences by whether or not they answered to
the gender section. Households of women who answered to the gender section appear to be
larger, have less assets in 2008 and less land. We include all these variables as controls in the
estimations we further present.
Our sample focuses on women who are in a union, which accounts for 268 women. How-
ever, we do not have information on all variables referring to empowerment and domestic
violence for all women in a union. Table 1 provides the number of observations based on the
non-missing data on empowerment and domestic violence. The final sample of women who
8In some cases, the eligible woman to answer the module was not present at the moment of the survey
8

are in a union and for which there is information on all variables referring to empowerment
and domestic violence, corresponds to 204 observations.
Before describing the factor analysis, we start motivating our analysis by building aggre-
gate measures of women’s empowerment and domestic violence. For this, we build one index
for women’s empowerment and one for domestic violence. To build the indices, we computed
the z scores for each of the variables used in each of the factors; and then, we computed
the mean of the z scores of the variables related to women’s empowerment and domestic
violence, separately. The top graph of Figure 1 shows how the index of women’s empow-
erment correlates to the index of domestic violence. The graph suggests that for low levels
of empowerment (eg. negative values), the relation between women’s empowerment and do-
mestic violence is flat; while for high levels of empowerment (eg. positive values), women’s
empowerment is positively correlated to domestic violence. The bottom graph shows the
Kernel distribution of the empowerment Index. To corroborate the relationship observed in
Figure 1, we estimate Equation 1.
IndexDVj=α+β1Indexj+β2Xj+j(1)
Where IndexDVjcorresponds to the index of domestic violence aggregating the aggres-
sive and the controlling ways for woman j. I ndexjcorresponds to the index aggregating all
measures of empowerment, including women’s income and education. Xjrepresents a set
of individual and household characteristics, such as household size, household head age and
education, assets owned in 2008, dummy for rural area, land size, woman’s age, woman’s age
at starting the union and previous experience of domestic violence, and a dummy indicating
whether partner and woman work. Finally, iis the error term.
Table 2 corroborates the results found in the graphical analysis. An increase of 1 stan-
dard deviations in the women’s empowerment index is associated with an increase of 0,3
standard deviations in the index of domestic violence. This is robust to adding controls. The
only household characteristic that significantly correlates to the index of domestic violence
is the dummy indicating whether or not women live in rural areas. Living in a rural area is
correlated with a 0,2 standard deviations increase in the likelihood of experience domestic
violence. Although this relation between women’s empowerment and domestic violence may
look a bit surprising, it goes in line with the economic theories that suggest that greater
domestic violence may also be a rational male response to the greater autonomy of women
[Bloch and Rao, 2002], [Eswaran and Malhotra, 2011]. Men use violence to increase their
bargaining power.
To better analyze how women’s empowerment correlate to domestic violence. We imple-
ment a factor analysis, following [Attanasio et al., 2015]. A latent factor model is usually
done to investigate concepts that are not easily and directly measured. In this case, different
observed variables are associated with women’s empowerment and domestic violence, which
9

corresponds to the two variables we are interested in describing. The factor analysis allows
understanding the patterns across the different observed variables, identifying measures (eg.
latent variables) representing a group of variables. Each factor captures a certain amount
of the overall variance in the observed variables; being the first factor the one that helps
explaining most of the variation.
To estimate the factor, we use the regression scoring (Thomson (1951)), which is defined
as: ˆ
f= Φ ∧0Σ−1x. Where, Φis the correlation matrix of the common factors, ∧represents
the factor loading matrix, and xthe vector of observed variables.. The loadings indicate
how important is the variable within the factor. Measures are allowed to load on more than
one factor, instead of assuming a dedicated measurement system. One of the purposes of
the analysis is also to see if one variable falls into more than one dimensions, or measures,
of empowerment. We also use an oblique rotation, as the different measures of women’s
empowerment and domestic violence can be correlated among them.
For the women’s empowerment factor analysis, we took into account variables referring to
self-esteem, disagreement towards domestic violence, willingness to divorce, intra-household
decision making, role in the productive organization receiving Oportunidades Rurales, and
the women’s social network. We consider that these variables attempt to cover different di-
mensions of women’s empowerment. Besides the variables included in the gender section, we
also include variables captured in the social capital section. The inclusion of the variables
related to social capital is not only motivated by the nature of our sample, eg. women and
men who participate in productive organizations; but also by previous evidence that have
suggested that women’s participation in micro-finance programs can challenge gender roles,
[Schuler et al., 1998]. A similar situation can take place with the participation in productive
organizations.
Variables related to self-esteem tell us about how confident women are of themselves,
as well as the variable related to willingness to get divorce9. We may think that more
self-confident women can exercise better agency in other domains, such as participating in
decision making or in productive organizations. A first step towards being more empowered,
can be acquiring higher self-esteem and confidence. In the same line, women who do not
agree with domestic violence may be more likely to stand for their rights. Participation
in household decision making has been a common, though not perfect, way to approach
women’s bargaining power, [Peterman et al., 2015], [Almas et al., 2015]. We may think that
more empowered women are more likely to participate in household’s decisions, such as in-
come use or children’s education. We include these variables in our analysis not only to see
how empowered women are in this dimension, but also to analyze whether this usual way to
approach women’s empowerment does help us to explain domestic violence. Our attempt is
to analyze these different dimensions of women’s empowerment, and see which ones help us
better explaining the experience of domestic violence.
9Table A.3 shows that the main reasons women reported to have thought about getting divorced are
partner’s infidelity (10%), aggressive behavior from the partner (8%) and partner’s alcohol consumption
(7%).
10

Table 3 shows the descriptive statistics of each of the measures used in the factor anal-
ysis. Around 50% of women expressed to be satisfied with themselves, and 58% strongly
disagrees about the use of domestic violence. Interestingly, only 18% of women strongly
disagrees with the statement that a good wife has always to obey her partner. This latter
figure may be suggesting that in rural settings, the male dominant role in the household is
still a social norm. Regarding intra-household decision making, more than 80% of women
say they participate in household decisions. The majority of women actively participate in
the productive organization, 38% are leaders, while 71% take part in the decision process.
Also, a large share of women is a source of information for other members of the community
and of the organization. We also include a variable indicating whether or not women have
thought about getting a divorce, 22% of our sample have thought about divorcing.
Table 4 shows the rotated factor loadings for the different measures of women’s empow-
erment. We worked with the four factors which eigenvalues were strictly larger than one10 .
The first factor loads mainly on the measures characterizing women’s self-esteem and dis-
agreement towards domestic violence (rows a to e and r). Factor 2 loads mainly on those
variables describing intra-household decision making (rows f to k). Factor 3 loads primarily
on the variables characterizing women’s role in the organization (rows l to n). Finally, factor 4
loads on variables describing whether the woman is a source of information in the community
(rows o to q). Each of these factors describes a different dimension of women’s empowerment,
allowing us to explore the sensitivity of the relationship with domestic violence.
To provide a general overview of how empowered women are in each of the dimensions
found through the factor analysis, Figures 2, 3, 4 and 5 show the Kernel distributions of
each factor of empowerment. Figures 2 and 4 show that in the dimensions of self-esteem
and role in the productive organization, women are relatively evenly distributed. Regarding
the dimension of participation in household decision making (Figure 3), there is very litte
variation and the majority of women have an average level of empowerment. While, for the
dimension of social network (Figure 4), the majority of women have relatively low levels of
empowerment.
The second factor analysis corresponds to the one on domestic violence. Table 5 shows
the descriptive statistics for the different episodes of domestic violence that were asked in the
survey. Based on figures from our sample, approximately 20% of women have experienced
episodes of emotional violence, such as being accused of being unfaithful, being humiliated
or being controlled by their partners. Around 10% of women have experienced physical vio-
lence, or have been threatened of physical or psychological abuse. Only 3% of women have
experienced sexual violence. According to the Demographic and Health Survey (DHS) of
2010 in Colombia, [Ojeda et al., 2011], around 32% of women reported to have received some
kind of threat from their partners. Threat to leave them for another woman, to take away
their children or to take away economic support. This figure is higher the older women are,
being 27% for women 15 to 19 years of age, and 34.3% for women 45 to 49 years of age.
10Factor 1 and 2 had an eigenvalue of 3 and explained 17% of the variation. Factor 3 had an eigenvalue
of 1.8 and explained 10% of the variation. And Factor 4 had an eigenvalue of 1.5 and explained 8% of the
variation.
11

Regarding physical violence, 37.4% of women have experienced any type of physical violence,
and 9.7% have experienced sexual violence. As in the case of threat of violence, rates increase
the older women are. For sexual violence, the rate is 5% for women 15-19 years of age, while
it is 13% for women 45 to 49 years of age. Finally, around 65% of women have experienced
some type of control behavior from their partners, such as accusing them of being unfaithful,
limiting contact with family and friends, ignoring them, or not taking them into account for
meetings and household decisions.
We consider that the rates of domestic violence in our sample can be lower, as compared
to the latest figures from the DHS in Colombia, because of the following reasons. First, our
sample only includes 6 out of 32 departments in Colombia11, plus we are working with a
sample of households that belong to productive organizations. This brings some selection
issues that need to be ackowledged. If we look at the rates of domestic violence reported at
the DHS, only for the departments we have in our sample, the rates of threat of violence,
physical violence and sexual violence are not particularly different from the national figures12.
This may be suggesting that selection is more an issue of women’s characteristics. Second,
when working with this kind of sensitive topics, issues of under-reporting and measurement
error can arise. Finally, for the particular case of physical violence, the DHS survey includes
a much larger list of episodes as compared to our list13. Despite the differences, previous
evidence reveals that women in Colombia, are highly exposed to domestic violence, physical
and emotional. Rates of physical violence in Colombia are among the largest in the region.
The highest rate of physical violence is observed in Bolivia (53%), followed by Colombia and
Peru (39%) [Bott et al., 2012]. These figures point out that domestic violence is an important
issue that can endanger women’s well-being and development.
Table 6 shows the rotated factor loadings for the different ways of domestic violence. We
worked with the two factors which eigenvalues were larger than one14. Factor 1 loads essen-
tially on the aggressive and physical ways of domestic violence, such as being humiliated,
being threatened physically and psychological, and being actually hit (rows d to l). The
second factor loads mostly on controlling ways, such as being accused of infidelity and being
controlled by their partners (rows a to c).
As we previously stated, the aim of this paper is to analyze the relationship between
women’s empowerment and domestic violence. For this, we not only use different measures
of women’s empworment and domestic violence, based on the factor analysis previously done.
We also include a set of household and individual characteristics that the theories, as well
as the empirical evidence, have shown to be correlated with domestic violence. Sociological
11Boyaca, Caldas, Cundinamarca, Huila, Risaralda and Tolima
12The rate of threat of violence in the 6 municipalities is 32%, of physical violence 38% and of sexual
violence 11%.
13We decided not to include as many episodes of physical violence because on one hand, we wanted to
have enough time for the other questions asked in the section. On the other hand, after discussing with the
research team, we considered that including many episodes of physical violence could make women feeling
uncomfortable.
14Factor 1 had an eigenvalue of 4.3 and explained 48% of the variation. Factor 2 had an eigenvalue of 1.24
and explained 14% of the variaton
12

theories have pointed out that women’s and partners’ resources are important at determin-
ing domestic violence, suggesting that women with higher status in the household are at a
higher risk of abuse. This set of theories also suggests that men facing an economic crisis or
unemployment, may be more vulnerable to perpetrate abuse [Goode, 1971]. The empirical
evidence is mixed and highly inconsistent regarding the relation between women’s labor force
participation and domestic violence. Some studies find that women’s labor market participa-
tion is positively associated with domestic violence, while others find a relation in the oposite
way [Kishor and Johnson, 2004], [Heisse, 2011]. Based on this set of theories we decided to
include a dummy indicating whether or not the women and their partners participate in
the labor market. From the psychological side, exposure to violence and poor parenting in
childhood can influence the likelihood of future violence. This influence can be through mod-
eling or through the negative impacts on child development [Dutton, 1995]. [Pollak, 2002]
proposed an inter-generational model of domestic violence. One of the assumptions of this
model is that individuals who grew up in violent homes tend to marry individuals who grew
up in violent homes. Motivated by this evidence, we include a dummy indicating whether or
not women witnessed domestic violence during childhood.
Last report from the Demographic and Health Survey for Colombia (2010) [Ojeda et al.,
2011], shows that women living in de-facto unions or in rural areas are more exposed to
domestic violence. The likelihood of having experienced any type of controlling behavior
from the partner, such as limiting the contact with family and friends or accuse her of being
unfaithful, is 64% for women in de-facto unions, 54% for married women, 67% for women
living in rural areas and 60% for women living in urban ones. Although the laws of many
countries recognize de-facto unions, civil and agrarian law does not seem to provide the same
benefits to women in this type of union15. Therefore, we may think that outside options
for women in a de facto union are not the same as for married women. This could make
them more vulnerable to experience domestic violence when partners consider their position
in the household is threatened [Kishor and Johnson, 2004]. Based on this, we also include
in our analysis a variable that indicates whether a woman is in a de-facto union. Follow-
ing this idea that men exercise violence when their position in the household is threatened
[Macmillan and Gartner, 1999], [Jewkes, 2002]; and use violence to leverage the bargaining
power within the househld [Bloch and Rao, 2002], [Eswaran and Malhotra, 2011]. We also
include a dummy for rural areas, where traditional gender roles seem to be more present.
One set of economic theories predict that an increase in women’s reservation utility leads
to a decrease in domestic violence [Farmer and Tiefenthaler, 1997], [Tauchen et al., 1991].
These models predict an decrease in domestic violence as women’s income increases (eg. part-
ners can buy less violence). [Tauchen et al., 1991] predicts that an increase in the woman’s
income is going to lead to a decrease in violence, while a rise in the husband’s income leads
to an increase. Having this in mind, we include in our analysis women’s and partner’s income
and education.
In different cross-sectional studies, evidence has shown a positive and significant correla-
15Taken from http://www.fao.org/docrep/u5615e/u5615e03.htm#P457_72830 on June 12th, 2015
13

tion between number of children and risk of physical violence. Some authors have interpreted
these results by suggesting that women with young children may be more willing to stay in a
violent relation because they are afraid they cannot raise their children alone [Heisse, 2011].
Based on this, we include not only the number of children within the household, but also
the total household size. We can think that in larger households domestic violence may be
less likely to happen because there can be more witnesses. We also include women’s and
partner’s age, as well as age at starting the union. Empirical evidence has shown that in
relations where age gaps are larger domestic violence is more likely to occur [Heisse, 2011].
Finally, we include the household land size as well as the assets they have in 2008 (eg. before
receiveing Oportunidades Rurales), as proxies for household’s wealth.
Before characterizing the relationship between women’s empowerment and domestic vio-
lence. We provide some descriptive analysis based on some of the households and individual
characteristics listed above. Table 7 shows that women in rural areas are indeed more likely
to experience domestic violence. Women in rural areas are more likely to be humiliated, to
be treated bad when they ask for money, and to be forced to do things they did not want
in the intimacy. 20% of women has been humiliated by their partners in rural settings, as
compared to 10% in urban ones. In households where the partner and the woman work,
women are less exposed to a controlling behavior from their partners. Particularly, they are
less exposed to limited contact with family and friends, as well as to being constantly asked
about where they were and with whom. Finally, women who have been witness of domestic
violence during childhood are more exposed to have been forced to do things they did not
want in the intimacy. For many of the other characteristics we do not observe significant
differences. Notably for women’s education, a variable commonly used to measure bargain-
ing power, we find no significant correlations. These patterns go in line with some, but not
certainly all of, previous empirical evidence; and their underlying economic and sociological
theories. [Pollak, 2002], [Goode, 1971].
3 Results
3.1 Main Results
Table 7 shows that some of these variables do correlate to certain questions on domestic
violence, but certainly not in a very systematic way.
To characterize the relation between women’s empowerment and domestic violence based
on the different latent measures we found through our factor analysis, we estimate Equation 2.
DVij =α+β1F1j+β2F2j+β3F3j+β4F4j+β5Xj+j(2)
Where DVij corresponds to the measure of domestic violence given by factor i (i=1,2) for
woman j. F ijcorresponds to the measure of empowerment given by factor i (i=1,2,3,4) for
woman j. Xjrepresents a set of individual and household characteristics, such as household
14

size, household head age and education, assets owned in 2008, dummy for rural area, land
size, woman’s age, education and income, woman’s age at starting the union and previous
experience of domestic violence, and a dummy indicating whether partner and woman work.
Finally, iis the error term.
Table 8 shows the results for aggressive ways of domestic violence (Factor 1) and Table 9
shows the results for controlling behavior from the partner (Factor 2) . In both tables, the
first column corresponds to the specification with all the four factors of empowerment; the
second column to the specification including the four factors of empowerment and woman’s
income and education; the third column to the specification with all individual and house-
hold characteristics except woman’s income and education; and the fourth column to the
specification with the whole set of individual and household characteristics (Xj). Table 8
shows that Factor 2 and 4 of empowerment are not significantly correlated to domestic vi-
olence. Factor 1 and 3 are positively correlated, and magnitudes are similar. An increase
of 1 standard deviations in women’s empowerment, proxied by their role in the productive
organization, is associated with a 0,15 standard deviations increase in the likelihood of ex-
periencing aggressive domestic violence; such as being humiliated or being hit or pushed by
their partners. An increase of 1 standard deviations in women’s empowerment, proxied by
self-esteem, disagreement towards domestic violence and willingness to get divorced, is asso-
ciated with a 0,13 standard deviations increase in aggressive domestic violence.
We do not find that reported participation in household decision making significantly
correlates to domestic violence. We may think that women’s participation in intra-household
decisions may not be accurately measured, as it is usually asked. On one hand, the decisions
spheres that are usually asked do not necessarily precisely capture women’s autonomy. Per-
haps, participation in decisions on how to invest family savings, or whether or not to sell a
family asset can tell more about women’s decision power. On the other hand, it is difficult
to establish whether sole or joint, with the partner, decision making implies more bargaining
power. As we saw, the majority of women take jointly decisions with their partners. This
latter set of results go in line with what has been found by [Almas et al., 2015], who have
shown that most women report to take decisions jointly with their partners, leaving limited
variation to observe changes in empowerment, proxied by household decision making. And
[Peterman et al., 2015], who show that women’s status, proxied by household decision mak-
ing, changes according to how the question is framed.
Regarding the individual and household characteristics included in our analysis, a larger
household size is correlated with lower levels of domestic violence, while a larger number
of children is positively correlated. Empirical evidence has shown that women living with
children can be more prone to violence, because they are more afraid of leaving the relation-
ship. The presence of one more child in the household is associated with an increase of 0,13
standard deviations in experiencing domestic violence. Women living in rural areas appear
to be more subject to abuse. Living in a rural area is correlated with an increase of 0,3
standard deviations in experiencing aggressive ways of domestic violence. Contrary to what
previous studies have found, we do not find that having experienced or witnessed violence as
a child is positively and significantly correlated to domestic violence. We may think that the
15

variable may not be accurately measured as there may be lack of recalling, or that we are
not necessarily capturing the severity of violence experienced.
Reading previous results can raise the question of what is the role played by women’s
income and education, which are the common and usual ways to approach women’s empow-
erment and/or women’s reservation utility. We consider that women’s education and income
are not only proxies of women’s empowerment, but of many other household and individual
characteristics. Results of Table 8 shows that although they are not statistically signifi-
cant, women’s income correlates positively with domestic violence, while women’s education
correlates negatively. This suggests that they can correlate with domestic violence through
different channels, rather than only women’s empowerment. Therefore, we decide to use
them as covariates instead of including them directly in the factor analysis. In Section 3.2
we provide additional estimations to support this.
Table 9 shows the results for controlling ways of domestic violence, such as being accused
of infidelity or being controlled. Contrary to what we found in Table 8, none of the proxies of
women’s empowerment help explaining partner’s controlling behavior. Being in a household
where both, women and partner, work is associated with a lower exposure to controlling
behavior. We can think that partners of women who participate in the labor market are
more used to women who go out of the home often, and meet other people; therefore use less
controlling behaviors. The participation of the partner and the woman in the labor market,
reduces the likelihood of begin controlled in 0,3 standard deviations. When we look at the R2
statistic, we also see that women’s empowerment does not explain controlling episodes of do-
mestic violence. The R2 is 1% as compared to 6% for aggressive ways of domestic violence.
Household and individual characteristics appear to contribute much more at explaining a
controlling behavior from the partners. On the other hand, a women’s social capital and self-
esteem appears particularly predictive of aggressive violence, and explains about as much
variation as the other covariates combined.
Results found in this paper go in line with the economic models that predict an increase
in domestic violence due to an increase in women’s bargaining power. Still, this is only true
when we analyze aggressive episodes of domestic violence. The motives that drive exercising
control over the women seem to be different, and more associated to household and individual
characteristics. As for the case of aggressive ways of domestic violence, it seems that partners
use violence as a way to leverage bargaining power within the household.
3.2 Robustness checks
We also estimate an equation similar to Equation 2 but including women’s income and educa-
tion inside the factor analysis. Table A.4 shows the results from the factor analysis including
women’s income and education. Results show that income and education load on different
factors. Women’s income loads on factor 1 and 3 (self-esteem and role in the organization),
and women’s education on factor 2 and 3 (Intra-household decision making and role in the
organization). In this factor analysis we also included women’s participation in other or-
16

ganizations, which loads on factors 3 and 4. Table A.5 shows similar results to what we
found in Table 8. Women’s empowerment, approached by women’s role in the organization
together with income; and self-esteem together with education, are the two dimensions of
empowerment positively and significantly correlated to the experience of aggressive episodes
of domestic violence.
The factors of women’s empowerment can be correlated among each other16. A woman
who is more self-determined may me more likely to participate in household decision making
or to engage in community organizations. To further validate that the factors of empower-
ment that significantly correlate to domestic violence are women’s self-determination (Factor
1) and role in the organization (Factor 3); we also estimate Equation 2 excluding factor 1
and 3 (Table A.6), and including only factor 1 and 3 of empowerment (Table A.7). Table A.6
shows that even when factors 1 and 3 are excluded, factors 2 and 4 do not seem to substan-
tially contribute to explaining domestic violence. When only factor 2 and 4 are included, they
only contribute 1% to explaining the variation in aggressive episodes of domestic violence;
while factors 1 and 3 contribute 5%.
4 Conclusions
Experiencing domestic violence has adverse health and socio-economic consequences, not only
for the women that experience the violence, but also for the children that witness it. Several
empirical studies have tried to understand the socio-economic characteristics that determine
being victim of abuse. Women who live in rural areas, who are in a de-facto union, who have
experience violence as a child and who have a larger number of children appear to be more
exposed to domestic violence. Women’s resources within the marriage play also an important
role. Women with higher education as compared to their partners, appear to be more ex-
posed to violence. Economic theory has modeled domestic violence through non-cooperative
bargaining models. One set of models have suggested that an increase in women’s reservation
utility will generate a decrease in domestic violence; while others have suggested that men
use domestic violence to leverage the bargaining power in the household, case in which an
increase in women’s reservation utility can lead to an increase in domestic violence.
This paper aims to contribute to the literature by providing descriptive evidence on the
relation between women’s empowerment and domestic violence. We make use of a compre-
hensive gender section designed for this purpose, in which we asked questions on different
dimensions of women’s empowerment, as well as on different events of domestic violence.
This extensive set of information allows us to test how responsive is the relation between
women’s empowerment and domestic violence, based on the way each of them is measured.
We implemented a factor analysis to get the different measures of women’s empowerment and
domestic violence. Our factor analysis suggests four measures for women’s empowerment,
one for women’s self-determination, one for intra-household decision making, one for the role
of women in productive organizations, and one last factor for women’s social network. For
16We performed an oblique rotation in our factor analysis having that in mind.
17

domestic violence, our factor analysis suggests a first factor for the controlling behavior from
the partner, and a second factor for the more aggressive ways of violence.
Our results suggest that the two factors of empowerment that better help explaining ag-
gressive episodes of domestic violence, are the factor describing women’s self-determination
and the factor describing women’s role in a productive organization. None of the dimensions
of empowerment we work with, appears to correlate to episodes of controlling behavior from
the partner. Socio-demographic characteristics such as living in a rural area and a larger
presence of children in the household, are also positively correlated to the experience of ag-
gressive situations of domestic violence.
Results found in this paper go in line with the economic models that predict an increase in
domestic violence due to an increase in women’s reservation utility. In these cases, partners
use violence as a way to leverage bargaining power within the household. We can think that
women, who are more self-determined, stand for their rights, are leaders in the productive
organizations and take part in decision process, are women who can be seen as a threat for
the traditional male dominance in the household. The setting in which we are working is a
setting mainly characterize by rural population, where we can think that male dominance is
more common. Contrary to what we initially would have expected, reported women’s par-
ticipation in intra-household decisions does not appear to significantly correlate to domestic
violence. We interpret this latter result by suggesting that the way the participation in
households decisions is usually asked may not be the most accurate way to capture women’s
autonomy.
To better understand domestic violence and design better social policies, much more
analysis is needed. Still, based on our results, we can highlight some important points. We
can suggest that including questions referring to women’s self-esteem and social capital is
important at trying to characterize women’s bargaining power. Also, it seems that further
analysis is needed to better capture women’s decision power in the household; as well as the
community characteristics that correlate with domestic violence. Besides, it appears that the
correlates of physical and aggressive violence are quite different, as compared to the correlates
of partner’s controlling behavior. This suggests that they should not be pooled together at
the moment to model domestic violence.
Finally, while the results in this paper are clearly not causal, the positive relationship
between different measures of women’s participation in different organizations and aggres-
sive domestic violence, nevertheless provide some reason to pause. Several social programs
encourage women to participate in meetings or promote the formation of women’s groups.
Despite the fact this kind of initiatives are thought to empower women and generate better
outcomes for them and their families, policy makers need to acknowledge the un-intended
consequences these interventions may have. Economic and sociological theories have pointed
out that men can use violence when their dominant role in the household is threatened. A
higher participation of women in meetings, or organizations, may constitute a behavior part-
ners does not like or consider menacing; using domestic violence as a way to reaffirm their
power within the household.
18

Tables
Table 1: Sample Size
Obs
Women who are in a union 268
Women for which I have info on domestic violence and are in a union 267
Women for which I have info on all proxies of empowerment, domestic violence and
are in a union
207
Women for which I have info on all proxies of empowerment, domestic violence, HH
and individual characteristics and are in a union
204
19

Table 2: correlates of Index of Domestic Violence
(1) (2)
Empowerment Index 0.298** 0.386***
(0.122) (0.136)
HH size -0.018
(0.03)
Partners’ age 0.004
(0.006)
Assets in 2008 (#) 0.019
(0.023)
Partner’s Educa (Yrs) 0.003
(0.011)
Rural area 0.193*
(0.1)
Land size (Hs) -0.002
(0.007)
Members under 18 y.o 0.046
(0.04)
Partner’s Income (Ln) 0.004
(0.018)
Woman’s age 0.005
(0.007)
Age at starting the union -0.001
(0.008)
Member 0.08
(0.079)
Experience violence as a child 0.131
(0.107)
De-facto union 0.076
(0.11)
Woman and partner work -0.128
(0.088)
R-square 0.028 0.092
Obs 207 204
Mean 0.000
Note: Standard errors in parenthesis and clustered at the organization level. *** p<0.001 ** p<0.05 * p<0.1.
Factor 1 refers to self-esteem and disagreement to domestic violence. Factor 2 describes intra-household
decision. Factor 3 characterize women’s role in the organization. Factor 4 describes women’s social network.
20

Table 3: Descriptive Statistics: Measures of Women’s Empowerment
Mean Obs
Are you satisfied with yourself? (strongly agrees) 0.480 204
You can do things as good as others (strongly agrees) 0.495 204
A good wife has always to obey her husband? (stronly disagrees) 0.181 204
Men have the right to hit their wives (strongly disagrees) 0.588 204
A woman has to stand being maltreated by her partner (stronly disagrees) 0.618 204
Who takes decisions on type of food: only she/she and her partner 0.936 204
Who takes decisions on clothing expenditures: only she/she and her partner 0.887 204
Who takes decisions on education expenditures: only she/she and her partner 0.882 204
Who takes decisions on how to use HH income: only she/she and her partner 0.868 204
Who takes decisions on how to use her income: only she/she and her partner 0.971 204
Who takes decisions on contraceptive methods: only she/she and her partner 0.926 204
Leader in the orga 0.368 204
Takes part in the decision process 0.716 204
Attend always to the meetings 0.515 204
Was asked for help by a member of the organization 0.186 204
Was asked for help by the leader of the organization 0.074 204
Was asked for help by other 0.363 204
Willingness to get divorced 0.216 204
Note: Descriptives for the final sample of women I work with (eg. no missing value in any of the variables used in the regression.)
21

Table 4: Factor Analysis. Women’s Empowerment
Factor 1 Factor 2 Factor 3 Factor 4
a Are you satisfied with yourself? (strongly agrees) 0.83 0.03 0.03 0.08
b You can do things as good as others (strongly agrees) 0.82 0.02 -0.01 0.08
c A good wife has always to obey her husband? (strongly disagrees) 0.53 0.00 0.28 -0.16
d Men have the right to hit their wives (strongly disagrees) 0.74 0.01 0.01 -0.18
e A woman has to stand being maltreated by her partner (strongly disagrees) 0.77 -0.08 -0.07 -0.18
f Decision maker on type of food: only she/she and her partner -0.08 0.60 -0.16 -0.11
g Decision maker on clothing expenditures: only she/she and her partner -0.08 0.69 -0.02 0.10
h Decision maker on education expenditures: only she/she and her partner 0.11 0.73 0.13 0.11
i Decision maker for to use HH income: only she/she and her partner 0.00 0.70 0.02 0.07
j Decision maker for on how to use her income: only she/she and her partner -0.04 0.79 0.02 -0.03
k Decision maker for on contraceptive methods: only she/she and her partner 0.08 0.63 0.07 -0.04
l Leader in the organization 0.16 0.06 0.72 0.12
m Takes part in the decision process -0.01 0.09 0.80 0.00
n Attends always to the meetings -0.16 -0.09 0.62 -0.10
o Was asked for help by a member of the organization -0.26 0.05 0.02 0.81
p Was asked for help by the leader of the organization 0.16 0.06 -0.09 0.70
q Was asked for help by other -0.05 0.01 0.21 0.56
r Has thought about getting divorced 0.31 0.11 0.14 0.13
Note: Oblique rotation
22

Table 5: Descriptive Statistics: Domestic Violence
Mean Obs
Has accused of being unfaithful 0.172 204
Doesn’t allow to meet family and friends 0.098 204
Has insisted in knowing where you are all the time 0.157 204
Has humiliated you 0.176 204
Has threatened to abandon you for another woman 0.098 204
Has treated you bad when you ask for money 0.088 204
Has threatened to hit you 0.083 204
Has pushed you or hit you 0.108 204
Has forced you to do things you didn’t want in the intimacy 0.049 204
Note: Descriptives for the final sample of women I work with (eg. no missing value in any of the variables
used in the regression.)
Table 6: Factor Analysis: Domestic Violence
Factor 1 Factor 2
a Has accused of being unfaithful 0.39 0.64
b Doesn’t allow to meet family and friends 0.30 0.79
c Has insisted in knowing where you are all the time 0.21 0.79
d Has humiliated you 0.71 0.38
e Has threatened to abandon you for another woman 0.66 -0.03
f Has treated you bad when you ask for money 0.76 0.27
g Has threatened to hit you 0.75 0.23
h Has pushed you or hit you 0.73 0.22
i Has forced you to do things you didn’t want in the intimacy 0.80 -0.16
Note: Oblique rotation
23

Table 7: Domestic Violence and HH Characteristics
Area Partner and
woman work
Urban Rural Diff (se) No Yes Diff (se)
Has accused of being unfaithful 0.125 0.193 (0.055) 0.190 0.164 (0.048)
Doesn’t allow to meet family and friends 0.078 0.107 (0.044) 0.172 0.068 (0.047)
Has insisted in knowing where you are all the time 0.156 0.157 (0.056) 0.241 0.123 (0.06)
Has humiliated you 0.109 0.207 (0.055) 0.190 0.171 (0.058)
Has threatened to abandon you for another woman 0.063 0.114 (0.041) 0.121 0.089 (0.051)
Has treated you bad when you ask for money 0.031 0.114 (0.038) 0.086 0.089 (0.042)
Has threatened to hit you 0.078 0.086 (0.049) 0.103 0.075 (0.043)
Has pushed you or hit you 0.063 0.129 (0.045) 0.103 0.110 (0.045)
Has forced you to do things you didn’t want in the intimacy 0.016 0.064 (0.028) 0.034 0.055 (0.029)
Obs 64 140 204 58 146 204
Experienced De facto union
viol as a child
No Yes Diff (se) No Yes Diff (se)
Has accused of being unfaithful 0.158 0.197 (0.048) 0.153 0.200 (0.065)
Doesn’t allow to meet family and friends 0.098 0.099 (0.038) 0.097 0.100 (0.051)
Has insisted in knowing where you are all the time 0.150 0.169 (0.051) 0.169 0.138 (0.052)
Has humiliated you 0.165 0.197 (0.062) 0.185 0.163 (0.055)
Has threatened to abandon you for another woman 0.090 0.113 (0.048) 0.097 0.100 (0.041)
Has treated you bad when you ask for money 0.083 0.099 (0.059) 0.089 0.088 (0.041)
Has threatened to hit you 0.053 0.141 (0.067) 0.073 0.100 (0.038)
Has pushed you or hit you 0.083 0.155 (0.061) 0.089 0.138 (0.041)
Has forced you to do things you didn’t want in the intimacy 0.023 0.099 (0.039) 0.056 0.038 (0.029)
Obs 133 71 204 124 80 204
At least 1 member Education higher
under 18 y.o than primary
No Yes Diff (se) No Yes Diff (se)
Has accused of being unfaithful 0.162 0.176 (0.058) 0.180 0.161 (0.039)
Doesn’t allow to meet family and friends 0.103 0.096 (0.05) 0.099 0.097 (0.037)
Has insisted in knowing where you are all the time 0.176 0.147 (0.063) 0.162 0.151 (0.048)
Has humiliated you 0.206 0.162 (0.049) 0.207 0.140 (0.048)
Has threatened to abandon you for another woman 0.059 0.118 (0.042) 0.117 0.075 (0.037)
Has treated you bad when you ask for money 0.118 0.074 (0.044) 0.099 0.075 (0.04)
Has threatened to hit you 0.059 0.096 (0.037) 0.081 0.086 (0.043)
Has pushed you or hit you 0.088 0.118 (0.04) 0.090 0.129 (0.042)
Has forced you to do things you didn’t want in the intimacy 0.059 0.044 (0.028) 0.054 0.043 (0.031)
Obs 68 136 204 111 93 204
Note: Descriptives for the final sample of women I work with (eg. no missing value in any of the variables used in the regression.). In bold the
differences that are statistically significant.
24

Table 8: Correlates of Aggressive Ways of Domestic Violence
(1) (2) (3) (4)
Empowerment (F1) 0.131** 0.130** 0.120* 0.124*
(0.065) (0.06) (0.064) (0.065)
Empowerment (F2) -0.014 -0.007 0.042 0.038
(0.043) (0.04) (0.04) (0.041)
Empowerment (F3) 0.151*** 0.148*** 0.156*** 0.153***
(0.051) (0.052) (0.055) (0.054)
Empowerment (F4) 0.097 0.099 0.111 0.111
(0.07) (0.072) (0.076) (0.076)
Woman’s income (Ln) 0.028 0.024
(0.029) (0.03)
Woman’s educa (yrs) -0.014 -0.017
(0.012) (0.017)
HH size -0.097** -0.095**
(0.041) (0.043)
Partners’ age -0.003 -0.002
(0.008) (0.008)
Assets in 2008 (#) 0.019 0.022
(0.042) (0.044)
Partner’s Educa (Yrs) 0.007 0.017
(0.017) (0.019)
Rural area 0.282* 0.282*
(0.152) (0.153)
Land size (Hs) 0.003 0.003
(0.012) (0.012)
Members under 18 y.o 0.129** 0.131**
(0.055) (0.055)
Partner’s Income (Ln) -0.005 -0.003
(0.024) (0.023)
Woman’s age 0.016 0.013
(0.01) (0.01)
Age at starting the union -0.006 -0.004
(0.012) (0.012)
Member 0.008 -0.003
(0.115) (0.119)
Experience violence as a child 0.264 0.257
(0.196) (0.2)
De-facto union 0.121 0.102
(0.14) (0.143)
Woman and partner work 0.007 -0.008
(0.125) (0.128)
R-square 0.06 0.07 0.12 0.13
Obs 207 205 204 204
Mean -0.082
Test F1=F3 (p-value) 0.76 0.78 0.65 0.72
Note: Standard errors in parenthesis and clustered at the organization level. *** p<0.001 ** p<0.05 * p<0.1.
Factor 1 refers to self-esteem and disagreement to domestic violence. Factor 2 describes intra-household
decision. Factor 3 characterize women’s role in the organization. Factor 4 describes women’s social network.
25

Table 9: Correlates of Domestic Violence: Controlling Behavior
(1) (2) (3) (4)
Empowerment (F1) 0.044 0.06 0.068 0.073
(0.058) (0.061) (0.067) (0.071)
Empowerment (F2) 0.024 0.025 0.041 0.031
(0.041) (0.046) (0.043) (0.047)
Empowerment (F3) 0.044 0.048 0.011 0.002
(0.074) (0.078) (0.083) (0.086)
Empowerment (F4) 0.085 0.064 0.097 0.094
(0.068) (0.071) (0.082) (0.083)
Woman’s income (Ln) 0.029 0.048*
(0.026) (0.028)
Woman’s educa (yrs) -0.016 -0.015
(0.014) (0.02)
HH size 0.108 0.115*
(0.07) (0.067)
Partners’ age 0.018 0.019
(0.012) (0.012)
Assets in 2008 (#) 0.026 0.03
(0.039) (0.041)
Partner’s Educa (Yrs) -0.001 0.008
(0.02) (0.027)
Rural area 0.047 0.051
(0.146) (0.142)
Land size (Hs) -0.01 -0.009
(0.008) (0.008)
Members under 18 y.o -0.097 -0.092
(0.091) (0.09)
Partner’s Income (Ln) -0.005 -0.001
(0.023) (0.023)
Woman’s age -0.015 -0.019
(0.014) (0.013)
Age at starting the union 0.017 0.019
(0.014) (0.015)
Member 0.179 0.156
(0.122) (0.126)
Experience violence as a child -0.078 -0.087
(0.13) (0.125)
De-facto union -0.021 -0.044
(0.157) (0.162)
Woman and partner work -0.343** -0.380**
(0.145) (0.15)
R-square 0.01 0.02 0.09 0.10
Obs 207 205 204 204
Mean -0.002
Note: Standard errors in parenthesis and clustered at the organization level. *** p<0.001 ** p<0.05 * p<0.1.
Factor 1 refers to self-esteem and disagreement to domestic violence. Factor 2 describes intra-household
decision. Factor 3 characterize women’s role in the organization. Factor 4 describes women’s social network.
26

References
M. Adato, B. de la Briere, D. Mindek, and A. Quisumbing. The impact of progresa on
women’s status and intrahousehold relations. International food policy research institute,
2000.
S. Alkire. Measuring agency: Issues and possibilities. Indian Journal of Human Development,
1, 2007.
S. Alkire, R. Meinzen-Dick, A. Peterman, A. Quisumbing, G. Seymour, and A. Vaz. The
women’s empowerment in agriculture index. OPHI working paper 58, 2013.
S. Alkire, E. Hogue, C. Kovarik, H. Malapit, R. Meinzen-Dickand, A. Quisumbing,
F. Ramzan, and K. Sproule. Measuring progress toward empowerment. women’s em-
powerment in agriculture index: baseline report. Feed the future. The US. Governement’s
Global Hunger and Food Security Initiative, 2014.
I. Almas, a. Armand, O. Attanasio, and P. Carneiro. Measuring and changing control:
Women’s empowerment and targeted transfers. NBER Working Paper 21717, 2015.
M. Angelucci. Love on the rocks: Domestic violence and alcohol abuse in rural mexico. The
B.E. Journal of Economic Analysis & Policy, 8, 2008.
O. Attanasio and A. Mesnard. The impact of a conditional cash transfer programme on
consumption in Colombia. Fiscal Studiesl, 27, 2006.
O. Attanasio, E. Battistin, and A. Mesnard. Food and cash transfers: Evidence from Colom-
bia. The Economic Journal, 122, 2012.
O. Attanasio, S. Cattan, E. Fitzsimons, C. Meghir, and M. Rubio-Codina. Estimating the
production function for human capital: results from a randomized control trial in Colombia.
NBER Working Paper, 20965, 2015.
F. Bloch and V. Rao. Terror as a bagaining instrument: a case study of dowry violence in
rural India. The American Economic Review, 92, 2002.
G. Bobonis, M. Gonzalez-Brenes, and R. Castro. Public transfers and domestic violence: The
roles of private information and spousal control. American Economic Journal: Economic
Policy, 5, 2013.
S. Bott, A. Guedes, M. Goodwin, and J. Adams. Violence against women in Latin America
and the Caribbean. a comparative analysis of population-based data from 12 countries.
Pan American Health Organization Report, 2012.
A. Camacho and C. Rodriguez. Who’s the boss at home after receiving conditional cash
transfers? Paper presented at the 2012 American Economic Association Annual meeting
in Chicago, Illinois, 2012.
27

D. Dutton. The domestic assault of women. Vancouver, University of British Columbia
Press, 1995.
Econometria. Informe de cierre del programa desarrollo de las oportunidades de inversion
y capitalizacion de los activos de las microempresas rurales en colombia. Econometria
Consultores, 2014.
M. Ellsberg, H. Jansen, L. Heise, C. Watts, and C. Garcia-Moreno. Intimate partner violences
and women’s physical and mental health in the WHO multi-country study on women’s
health and domestic violence: an observational study. The Lancet, 371, 2008.
M. Eswaran and N. Malhotra. Domestic violence in developing countries: Theory and evi-
dence. Canadian Journal of Economics, 2011.
A. Farmer and J. Tiefenthaler. An economic analysis of domestic violence. Review of Social
Economy, 1997.
D. Flake. Individual, family, and community risk makers for domestic violence in peru.
Violence against women, 11, 2005.
A. Gage. Women’s experience of intimate partner violence in haiti. Social Science & Medicine,
61, 2005.
W. Goode. Force and violence in the family. Annual Review of Sociology, 11, 1971.
L. Heisse. What works to prevent partner violence: an evidence overview. London: Strive
Research Consortium, 2011.
L. Heisse. Determinants of partner violence in low and middle income-countries: explor-
ing variation in individual and population-level risk. Doctoral Thesis. London School of
Hygiene Tropical Medicine, 2012.
T. Herrenkohl, C. Sousa, E. Tajima, R. Herrenkohl, and C. Moylan. Intersection of child
abuse and children’s exposure to domestic violence”. Trauma, Violence and Abuse, 9, 2008.
M. Hidrobo and L. Fernald. Cash transfers and domestic violence. Journal of Health Eco-
nomics, 32, 2013.
R. Jewkes. Intimate partner violence: causes and prevention. Lancet, 359, 2002.
N. Kabeer. Resources, agency, achievements: Reflections on the measurement of women’s
empowerment. Journal of Development and Change, 30, 1999.
J. Kim, C. Watts, J. Hargreaves, L. Ndhlovu, G. Phetia, L. Morison, J. Busza, J. Porter, and
P. Pronky. Understanding the impact of a microfinance-based intervention on women’s
empowerment and the reduction of intimate partner violence in south africa. American
Journal of Public Health, 97, 2007.
S. Kishor and K. Johnson. Profiling domestic violence. a multi-country study. Measure DHS,
2004.
28

D. Levinson. Violence in cross-cultural perspective. Newbury Park, California, Sage Pub-
lishers, 1989.
R. Macmillan and R. Gartner. When she brings home the bacon: Labor-force participation
and the risk of spousal violence against women. Journal of Marriage and the Family, 61,
1999.
A. Morrison and M. Orlando. Social and economic costs of domestic violence: Chile and
Nicaragua. In Too close to home. Domestic violence in the Americas, 1999.
G. Ojeda, M. Ordonez, and L. Ochoa. Encuesta nacional de demografia y salud. ENDS 2010.
2011.
A. Peterman, B. Schwab, S. Roy, M. Hidrobo, and D. Gilligan. Measuring women’s decision-
making. Ifpri discussion paper, 01453, 2015.
R. Pollak. An intergenerational model of domestic violence. NBER Working paper, 9099,
2002.
R. Ribero and F. Sanchez. Determinantes, efectos y costos de la violence intrafamiliar en
Colombia. Documento CEDE, 44, 2004.
S. Schuler, S. Hashemi, A. Riley, and S. Akhter. Credit programs, patriarchy and men’s
violence against women in rural Bangladesh. Soc. Sci. Med., 43, 1996.
S. Schuler, S. Hashemi, and S. Badal. Men’s violence against women in rural Bangladesh:
Undermined or exacerbated by microcredit programmes? Development in Practice, 8,
1998.
H. Tauchen, A Witte, and S. Long. Domestic violence: A nonrandom affair. International
Economic Review, 32, 1991.
Unicef. Hidden in plain sight. A statistical analysis of violence against children. UNICEF,
2014.
WB. Voice and agency: empowering women and girls for shared prosperity. World Bank,
2014.
29

Figures
30

Figure 1: Domestic Violence and Women’s Empowerment
Note: Sample of women who are un a union. Lowess Graph. Kernel Distribution
31

Figure 2: Distribution Factor 1 of Empowerment: Self-esteem/Disagreement to DV
Note: Sample of women who are in a union. Kernel Distribution
Figure 3: Distribution Factor 2 of Empowerment: Intra-HH Decision Making
Note: Sample of women who are in a union. Kernel Distribution
32

Figure 4: Distribution Factor 3 of Empowerment: Role in the Organization
Note: Sample of women who are in a union. Kernel Distribution
Figure 5: Distribution Factor 4 of Empowerment: Diffusion of Information
Note: Sample of women who are in a union. Lowess Graph. Kernel Distribution
33

A Appendix
Table A.1: Descriptive Statistics on the Whole Sample
Mean Obs
HH size 3.919 729
Number of women in the HH 1.857 729
HH head age 50.313 721
HH spouse age 45.475 554
HH head educa 5.610 715
HH spouse educa 6.421 553
Number of children 0-18 yo in the HH 1.021 729
Assets in 2008 (#) 1.053 729
Land size 6.401 729
In union 0.730 722
Answered to gender section 0.392 729
Table A.2: Descriptive Statistics by whether or not they answered to the gender section
Answered
(Yes)
Answered
(No)
Difference SE (Diff) Obs
HH size 4.122 3.788 0.335 (0.134) 729
Number of women in the HH 1.934 1.808 0.125 (0.084) 729
HH head age 49.454 50.872 -1.418 (0.987) 721
HH spouse age 45.948 45.025 0.924 (1.022) 554
HH head educa 5.740 5.525 0.215 (0.307) 715
HH spouse educa 6.478 6.368 0.109 (0.338) 553
Number of children 0-18 yo in the HH 1.087 0.977 0.110 (0.07) 729
Assets in 2008 (#) 0.836 1.194 -0.358 (0.134) 729
Land size 4.584 7.574 -2.990 (0.876) 729
In union 0.926 0.603 0.323 (0.032) 722
Note: Standard errors in parenthesis
34

Table A.3: Reasons to have thought about getting divorced
Rate
Partner does not meet his responsabilities 0.049
Partner’s alcohol consumption 0.071
Partner has been unfaithful 0.093
Partner has been aggressive to you 0.078
Partner has been aggressive to children 0.026
35

Table A.4: Factor Analysis. Women’s Empowerment-Including women’s income and education inside the factor analysis
Factor 1 Factor 2 Factor 3 Factor 4
Are you satisfied with yourself? (strongly agrees) 0.03 0.82 -0.01 0.09
You can do things as good as others (strongly agrees) 0.02 0.82 -0.05 0.08
A good wife has always to obey her husband? (stronly disagrees) 0.00 0.53 0.26 -0.16
Men have the right to hit their wives (strongly disagrees) 0.02 0.73 0.06 -0.21
A woman has to stand being maltreated by her partner (stronly disagrees) -0.07 0.75 -0.02 -0.21
Decision maker on type of food: only she/she and her partner 0.61 -0.08 -0.12 -0.10
Decision maker on clothing expenditures: only she/she and her partner 0.70 -0.09 -0.02 0.08
Decision maker on education expenditures: only she/she and her partner 0.73 0.10 0.15 0.13
Decision maker on how to use HH income: only she/she and her partner 0.70 -0.01 0.01 0.08
Decision maker on how to use her income: only she/she and her partner 0.82 0.00 0.01 -0.04
Decision maker on contraceptive methods: only she/she and her partner 0.63 0.10 0.03 -0.05
Leader in the orga 0.05 0.17 0.68 0.13
Takes part in the decision process 0.09 -0.02 0.77 0.04
Attend always to the meetings -0.09 -0.17 0.59 -0.19
Leader in other organizations -0.01 -0.04 0.42 0.41
Woman’s income (LN) 0.21 -0.11 0.26 -0.10
Woman’s education (years) 0.05 0.23 0.26 0.14
Was asked for help by a member of the organization 0.06 -0.25 0.04 0.78
Was asked for help by the leader of the organization 0.08 0.17 -0.15 0.58
Was asked for help by other 0.00 -0.03 0.21 0.58
Has thought about getting divorced 0.13 0.28 0.14 0.10
Note: Oblique rotation
36

Table A.5: Correlates of Agressive ways of Domestic Violence. Factor analysis including women’s
income and education
(1) (2)
Empowerment (F1) 0.131** 0.120*
(0.065) (0.064)
Empowerment (F2) -0.014 0.042
(0.043) (0.04)
Empowerment (F3) 0.151*** 0.156***
(0.051) (0.055)
Empowerment (F4) 0.097 0.111
(0.07) (0.076)
HH size -0.097**
(0.041)
Partners’ age -0.003
(0.008)
Assets in 2008 (#) 0.019
(0.042)
Partner’s Educa (Yrs) 0.007
(0.017)
Rural area 0.282*
(0.152)
Land size (Hs) 0.003
(0.012)
Members under 18 y.o 0.129**
(0.055)
Partner’s Income (Ln) -0.005
(0.024)
Woman’s age 0.016
(0.01)
Age at starting the union -0.006
(0.012)
Member 0.008
(0.115)
Experience violence as a child 0.264
(0.196)
De-facto union 0.121
(0.14)
Woman and partner work 0.007
(0.125)
R-square 0.06 0.12
Obs 207 204
Note: Standard errors in parenthesis and clustered at the organization level. *** p<0.001 ** p<0.05 *
p<0.1.
37

Table A.6: Correlates of Agressive ways of Domestic Violence. Excluding factor 1 and 3 of
empowerment.
(1) (2) (3)
Empowerment (F2) -0.005 0.054 0.049
(0.04) (0.043) (0.043)
Empowerment (F4) 0.098 0.109 0.108
(0.074) (0.078) (0.078)
Woman’s income (Ln) 0.029
(0.03)
Woman’s educa (yrs) -0.012
(0.017)
HH size -0.078* -0.075*
(0.039) (0.041)
Partners’ age -0.003 -0.002
(0.009) (0.009)
Assets in 2008 (#) 0.014 0.017
(0.043) (0.044)
Partner’s Educa (Yrs) 0.016 0.023
(0.017) (0.019)
Rural area 0.311** 0.312**
(0.147) (0.149)
Land size (Hs) 0.003 0.004
(0.012) (0.012)
Members under 18 y.o 0.089 0.092
(0.057) (0.056)
Partner’s Income (Ln) 0.009 0.011
(0.023) (0.023)
Woman’s age 0.014 0.012
(0.011) (0.011)
Age at starting the union -0.007 -0.006
(0.012) (0.012)
Member 0.137 0.119
(0.123) (0.13)
Experience violence as a child 0.295 0.29
(0.202) (0.205)
De-facto union 0.086 0.072
(0.132) (0.135)
Woman and partner work -0.008 -0.03
(0.12) (0.124)
R-square 0.011 0.082 0.088
Obs 207 204 204
Note: Standard errors in parenthesis and clustered at the organization level. *** p<0.001 ** p<0.05 *
p<0.1.
38

Table A.7: Correlates of Agressive ways of Domestic Violence. Including only factor 1 and 3 of
empowerment.
(1) (2) (3)
Empowerment (F1) 0.133** 0.120* 0.124*
(0.063) (0.062) (0.063)
Empowerment (F3) 0.149*** 0.158*** 0.154***
(0.051) (0.057) (0.056)
Woman’s income (Ln) 0.028
(0.03)
Woman’s educa (yrs) -0.015
(0.017)
HH size -0.094** -0.092**
(0.041) (0.042)
Partners’ age -0.004 -0.003
(0.008) (0.008)
Assets in 2008 (#) 0.017 0.021
(0.043) (0.045)
Partner’s Educa (Yrs) 0.007 0.016
(0.019) (0.02)
Rural area 0.265* 0.267*
(0.155) (0.157)
Land size (Hs) 0.003 0.003
(0.012) (0.012)
Members under 18 y.o 0.129** 0.132**
(0.053) (0.052)
Partner’s Income (Ln) -0.001 0.002
(0.025) (0.025)
Woman’s age 0.016 0.014
(0.01) (0.01)
Age at starting the union -0.008 -0.007
(0.011) (0.012)
Member 0.002 -0.01
(0.118) (0.122)
Experience violence as a child 0.245 0.239
(0.195) (0.2)
De-facto union 0.118 0.101
(0.144) (0.146)
Woman and partner work 0.03 0.01
(0.125) (0.129)
R-square 0.047 0.105 0.111
Obs 207 204 204
Note: Standard errors in parenthesis and clustered at the organization level. *** p<0.001 ** p<0.05 *
p<0.1.
39