36151




       The costs and impacts of gender-based violence in developing countries:

                    Methodological considerations and new evidence




                                      Andrew R. Morrison
                                     Maria Beatriz Orlando1




                                         November 2004




1We would like to thank Omar Arias, Marina Bassi, Myra Betron, Suzanne Duryea, Jonathon Goldberg,
Rosangela Leao, Vijayendra Rao, Laura Ripani, Cecilia Valdivieso and Bernice van Bronkhorst for their
comments and suggestions. All remaining errors, of course, are our own. Data for estimations were
obtained from ORC Macro's Measure DHS+ program. Estimations were performed using the Propensity
Score Matching program for STATA by Sascha Becker and Andrea Ichino.

Executive Summary


Gender-based violence (GBV) is undoubtedly one of the most common forms of violence
in the world. GBV includes physical violence, emotional violence, sexual violence and
economic violence. Recent multi-country studies, using a common methodology and
definitions, have found high prevalence of physical and sexual violence against women
by intimate male partners.

GBV has devastating consequences not only for victims but also for society as a whole. A
growing body of research has focused on documenting and dimensioning these
consequences either by estimating the impacts of GBV on a set of outcomes or by
measuring the monetary costs of GBV. Studies for developing countries conclude that
the health impact of GBV on women can be as high as some of the leading causes of
injury; consequences are especially serious in the area of reproductive health. Previous
studies estimating the socioeconomic costs of GBV have documented the impact of GBV
on earnings due to death and lost productivity, job loss, lost productivity of the abuser
due to incarceration, and loss of tax revenues due to death and incarceration.

This paper has three goals. First, it synthesizes the research on estimating the costs and
impacts of GBV, presenting and evaluating the methodologies used to produce cost and
impact estimates. While estimates are presented for both developed and developing
countries, the focus is on methodologies which are potentially most relevant to
developing countries. A second goal of this paper is to use a relatively new
methodology--propensity score matching--to generate comparable, cross-country
estimates of some of the impacts of GBV in Haiti, Peru and Zambia. This methodology
has not previously been used on a cross-country basis to estimate the impacts of GBV.
Finally, the paper identifies priority areas for future research.

Evaluation of      methodologies to estimate costs and impacts of GBV. All the
methodologies presented in this paper have advantages and disadvantages. Accounting
approaches to estimating the costs of GBV are straightforward to carry out, but may
seriously underestimate the true social costs of violence and may be especially
problematic in a developing country context. Indirect cost estimates of the impact of
GBV on women's labor force participation and earnings have been successful in
highlighting the productivity impacts of gender-based violence and have helped establish
that GBV is an important to both economic development and gender equality; at the same
time, indirect cost estimates have suffered from econometric weaknesses that have yet to
be definitively resolved. DALY measures of the health impacts of GBV have proved
indispensable for firmly placing the issue of GBV on the public health agenda, yet they
have not been particularly useful in terms of formulating policy responses to violence and
seem to have had a limited impact beyond the health sector. Naïve comparison of mean
outcomes between women who have suffered violence and those who have not have
contributed to identifying the specific health impacts on women and their children, but
have been criticized for being less than rigorous in their formation of control groups of
non-victims. Finally, propensity score matching adds rigor in the specification of control



                                                                                         2

groups, but does not solve simultaneity issues. In sum, there is no perfect methodology
with which to gauge the costs and impacts of GBV. All methodologies have strengths and
weaknesses, and the challenge is to choose the appropriate methodology given both data
constraints and the potential consumers of the estimates.

Estimates of health, employment, and education impacts of GBV in Haiti, Peru and
Zambia. We find strong impacts of GBV on women's health outcomes, especially on
reproductive health. GBV also increases--not surprisingly--women's usage of health
services. Lifetime physical violence by intimate partner increases the probability of
working in a paid job in all three countries. This result, of course, may be the result of
simultaneity between earnings and violence which is not resolved by PSM: women who
work outside the home threaten male supremacy and hence are more frequent victims of
violence. The study finds strong evidence that GBV may lead to the inter-generational
transmission of violence. Women who are victimized by physical violence--whether at
the hands of intimate partners or relatives--are more likely to use violence in disciplining
their children. Much scientific evidence has documented that children exposed to such
violence are themselves more likely to engage in violent and delinquent behaviors. The
study also finds an impact of GBV on children's health outcomes, with the specific
manifestations varying by type of GBV and by country. Where impacts are found on
children's educational performance, they are of an unexpected direction, with children of
women victims more likely to attend school and less likely to be behind in age/grade
progression.

Future research on the costs and impacts of GBV. Future research on the costs and
impacts of GBV has many promising fronts to explore. First, it would be useful to
employ the PSM methodology on richer data sets that include more information on
women's labor force participation and earnings, as well as household consumption
patterns.    This would allow the indirect costs of GBV to be explored in a
methodologically rigorous way. Second, more work needs to be done on the relationship
between income, wealth and violence. Any research has to deal with simultaneity issue,
which PSM does not resolve. Third, the results from this paper suggest that GBV is more
prevalent in urban areas--which is exactly opposite the conclusion emerging from the
WHO's multi-country study. Studies on this issue will need to address under-reporting of
violence, which is frequently assumed--without evidence--to be a larger problem in
rural areas; these studies necessarily will be qualitative in nature. Finally, there is a need
for more studies on the economic costs of regionally-specific types of GBV such as
female genital mutilation, violence generated by dowry systems, and incest. These
studies on economic costs of gender-based violence are essential to position the issue as
what it is: not only a woman's issue, a public health issue, and a human rights issue, but
also as an issue affecting poverty, development and economic growth.




                                                                                             3

1. Introduction

Gender-based violence (GBV) is undoubtedly one of the most common form of violence
in the world. GBV includes physical violence, emotional violence, sexual violence and
economic violence. Recent multi-country studies, using a common methodology and
definitions, have found high prevalence of physical and sexual violence against women
by intimate male partners, whether measured in the past 12 months or over women's
lifetimes. Other forms of GBV such as rape, trafficking, dowry-related violence and
female genital mutilation are also important and have been recently documented.2

An expanding body of research has attempted to document the impact that GBV has on
women and societies. There have been two strands of research. One has focused on
estimating the impact of GBV without imputing a monetary value to this impact. This
research has focused on the impact of GBV on women's reproductive health (unwanted
pregnancies, sexually-transmitted infections, and HIV/AIDS infection), mental health
(drug and alcohol abuse, depression, low self-esteem, and post-traumatic stress disorder)
and physical well-being (mortality and morbidity)--as well as on children's
developmental outcomes (low birth weight, diminished school performance, and violent
and delinquent behaviors).3 A second strand has attempted to estimate the costs to
society of GBV, using various methodological approaches.

This paper has three goals. First, it will synthesize the research on estimating the costs
and impacts of GBV, presenting and evaluating the methodologies used to produce cost
and impact estimates. While estimates will be presented for both developed and
developing countries, the focus will be on methodologies which are potentially most
relevant to developing countries. A common problem of some costing methodologies is
the difficulty of disentangling the impacts of GBV on health, employment, and education
outcomes from that of factors that increase the risk of GBV but also have a direct impact
on the outcomes of interest. Thus, a second goal of this paper is to use a relatively new
methodology--propensity score matching--that allows one to control for most
differences between women who have been abused and those who have not, except for
the occurrence of violence. Using this methodology, we generate comparable, cross-
country estimates of some of the impacts of GBV. This methodology has not previously
been used on a cross-country basis to estimate the impacts of GBV. Finally, the paper
will identify areas of priority research in the area of gender-based violence.

The following section provides a definition of gender-based violence, along with recent
estimates of its prevalence. Section 3 contains an analytical review of methodologies to
generate estimates of the socio-economic costs of                 GBV.         Section 4 presents our
estimates of the health, employment, and education impacts of GBV on abused women
and their children using propensity score matching. Finally, Section 5 offers some
conclusions and suggestions for future research.



2 See Bloch and Rao (2002) for a study of dowry violence in rural India.
3 See Heise, Ellsberg and Gottemoeller (1999) for a detailed discussion of the health consequences of
gender-based violence.


                                                                                                      4

2. Gender-based Violence: Definitions and Prevalence

The United Nations adopted the first internationally-accepted definition of violence
against women in 1993, defining such violence to be "any act of gender-based violence
that results in, or is likely to result in, physical, sexual or psychological harm or suffering
to women...whether occurring in public or private life (United Nations, 1993)." This
definition hints that gender-based violence is a broader category than violence against
women, and that gender-based violence--at least theoretically--could also include
violence that affects men, as long as it is conditioned upon or affected by men's gender
roles.

In practice, however, gender-based violence (GBV) has come to be almost synonymous
with violence against women.               The adjective "gender-based" is frequently used to
highlight the role that women's subordinate status in society plays in increasing the risk
that they will be impacted by violence (Heise et al., 1999; IPPF, 2000). Even if this
rough equivalence between violence against women and GBV is accepted, it is necessary
to more precisely define what types of violence should be included within GBV. A
comprehensive list of the types of GBV would include domestic violence, sexual abuse
and rape, trafficking of women, female genital mutilation, dowry violence, and other
forms of violence against women.4

Figure 1 presents data on physical abuse of women by intima`te partners. Data come
from the WHO multi-country study, which uses a standard methodology and identical
survey instruments across a number of developed and developing countries, thus
producing for the first time prevalence numbers which are comparable across countries.5
In urban areas rates of victimization in the last 12 months range from a low of 3% in
Serbia and Japan, to a high of 25 and 29% in provincial Peru and Ethiopia, respectively.
In terms of lifetime violence, victimization rates range from a low of 13% in Japan to
highs of 50 and 62% in Lima and provincial Peru, respectively. For countries in which
data are available for both the capital city and provincial areas, victimization rates are
almost invariably higher in provincial areas--both for violence in the last 12 months and
lifetime violence.6

Table 1 reports the percentage of women who reported suffering sexual violence at the
hands of a male intimate partner, either in the 12 months prior to the survey or at any
time in their life. Some of these data (Brazil, Japan, Peru and Thailand) come from the
WHO multi-country study on violence against women and are thus comparable across
these countries;7 the remaining data come from surveys using varying methodologies and

4This paper will provide data on the most widespread forms of gender-based violence: physical and sexual
violence by intimate partners, and sexual assault regardless of the identity of the aggressor.
5Japan was the sole developed country included in the multi-country study.
6Bangladesh is the sole exception, where violence over the last 12 months was higher in urban than in
provincial areas.
7Preliminary results on sexual violence by intimate partners from the WHO multi-country study are
available for only four of the ten countries for which data on physical violence by intimate partners are
currently available.


                                                                                                          5

thus cross-country comparisons are risky. Among the countries participating in the WHO
study, between 1.3% (Yokohama, Japan) and 22.9% (Cusco, Peru) of women reported
that male intimate partners had attempted or completed sexual violence in the preceding
12 months. The simple average for cities in the three developing countries (Brazil, Peru
and Thailand) is 11.85%.

Table 2 reports rates of victimization from sexual assault in the previous five years,
whether by intimate partners or others. Rates range from highs of 8% (Rio de Janeiro)
and 6% (Tirana) to 0.8% (Gaborone) and 0.3% (Manila).8 Given that the surveys did not
explicitly ask about sexual assault by intimate partners and that women frequently will
not classify sexual coercion by intimate partners as sexual assault, these numbers are
almost certainly underestimates. While strict comparisons with the data in Table 1 on
sexual victimization by intimate partners are not possible because there is little overlap in
terms of countries covered and no overlap in terms of cities covered, the rates of sexual
victimization by intimate partner in the last year in Table 1 are generally significantly
higher than the rates of sexual assault by any person over the previous past five years
reported in Table 2, lending credence to the assertion that the rates reported in Table 2 are
biased downward.


3. Methodological approaches to estimating the costs and impacts of GBV


This section will present methodologies for estimating the socio-economic costs and
impacts of GBV. Two types of methodologies are discussed: those that attempt to attach
a monetary value to the effects of GBV, and those that concentrate on estimating its
impacts on health, education, labor force participation and other outcomes of interest,
without attaching a monetary value. Table 3 summarizes the advantages and
disadvantages of the methodologies that will be discussed in the following subsections.

   A. Methodologies for the establishing the monetary costs of GBV

Before presenting the methodological options for producing cost estimates, a more
fundamental question must be addressed: why produce cost estimates at all? There are
two principle motivations. First, cost estimates can quantify the impact of GBV on
society and consequently influence the decisions of those who form public policy and
allocate funds; cost estimates, in other words, facilitate the "dimensioning" of the issue:
how important is GBV among the panoply of pressing development issues? Second,
cost estimates can help establish the potential benefits of violence prevention strategies or
programs (CDC, 2003; Miller, Cohen and Wiersema, 1996; Snively, 1994; Haddix et al,
1996; Teutcsch, 1992).




8Note that sexual assault includes both sexual assault by intimate partners and assault by non-partners.


                                                                                                         6

An accounting methodology for estimating direct costs

The most common approach used to calculate the costs of GBV has been an "accounting
methodology", in which costs are calculated for specific categories of costs, and total cost
to society is simply the sum of all distinct categories of costs. Typical of this approach is
CDC (2003), which specifies two types of costs:

    · Direct costs are actual expenditures related to GBV, including health care
        services, judicial services and social services.
    · Indirect costs represent the value of lost productivity from both paid work and
        unpaid work, as well as the foregone value of lifetime earnings for women who
        have died as a result of GBV.

Direct costs are typically calculated by establishing the unit cost of a specific service and
multiplying this unit cost by the number of time the service was utilized. Information on
women's use of medical and mental health services is usually gathered from population-
based surveys, and health care costs via a survey of medical service providers. (See Box
1 for the steps typically followed in estimating direct costs.)

Similar methodologies have been employed by to generate estimates for a number of
developed countries and states or provinces within these countries.9 There have been,
however, very few studies that have attempted to estimate the direct costs of GBV in
developing countries; to the best of our knowledge, there are only two such direct cost
studies for GBV in developing countries. Mansingh and Ramphal (1993) estimated that
the direct costs of treating victims of intimate partner violence in Kingston Public
Hospital (Jamaica) totaled US $454,000 in 1991 (in 2001 dollars). Sánchez et al. (2004)
find that the Colombian national government spent approximately 184 billion
pesos (US $73.7 million) in 2003 to prevent, detect and offer services to survivors of
intimate partner violence--an amount equal to approximately 0.6% of the total national
budget.10

Direct cost estimates are problematic, especially in a developing country context. A
long-accepted tenet of public finance economics is that the optimal level of public goods
provision occurs when the sum of the marginal willingness to pay for these services is
equal to the marginal cost of their provision. Actual expenditures may be a particularly
poor indicator as to whether the optimal amount of the public good is being provided. A
priori, there is a strong suspicion that services for victims of GBV are under-provided in


9Countries and states for which this type of analysis has been performed inlcude the U.S. (Laurence and
Spalter-Roth, 1995), Canada (Greaves et al., 1995; Health Canada, 2002), Australia (Australian Institute of
Criminology, 2002), British Columbia (Kerr and McLean, 1996), Holland (Korf et al.,1997), Northern
Territory (Office of Women´s Policy, 1996), Queensland (Blumel et al., 1993), Switzerland (Godenzi and
Yodanis, 1998) and the U.K. (Stanko et al., 1998). These studies are reviewed in Yodanis et al. (2000) and
WHO (2004).
10These represent expenditures of the following institutions: Colombian Social Welfare Insitute (Instituto
Colombiano de Bienestar Social), Sub-secretariat for Security and Peaceful Coexistence (Subsecretaria
para Asuntos de Seguridad y Convivencia Ciudadana), Attorney General´s Office (Fiscalía General de la
Nación) and the National Medical-Legal Institute (Instituto Nacional de Medicina Legal).


                                                                                                          7

Box 1. Steps in estimating the direct costs of intimate partner violence (IPV): the
CDC's approach for the U.S.

Step 1. Determine usage of medical services by women who were injured as a result of
IPV. Usage was calculated separately for rape and for physical assault. According to data
from the National Violence against Women Survey (NVAWS), 322,230 IPV rapes occur in
the U.S. each year. Of these, 36.2% result in physical injuries, of which 31% require
medical care. Also using NVAWS data, 4,450,807 IPV physical assaults occur annually, of
which 41.5% cause physical injuries. Medical care is sought in 28.1% of these cases

Step 2. Determine usage of mental health services by all women victimized by IPV,
whether or not physical injuries present. According to NVAWS data, 33% of all IPV
rapes result in usage of mental health services.

Step 3. Establish the unit cost of services used. Services whose costs were calculated
include: ambulance transport, paramedic care, emergency room care, physician visit,
physical therapy, dental visits, inpatient hospitalizations, outpatient clinic visits. The unit
costs were established by using the Medical Expenditure Panel Survey and the Medicare 5%
Sample Beneficiary Standard Analytic Files, and these unit costs were then multiplied by
service usage as revealed in the NVAWS.

The mean medical care costs per incident (including those incidents in which women did
not seek treatment) were:

         $516 per IPV rape
         $548 per IPV assault

The mean mental health care costs per incident (including those incidents in which women
did not seek treatment) were:
         $323 per IPV rape
         $269 per IPV physical assault
         $294 per IPV stalking

Step 4. Calculate total direct costs as the product of unit costs times the number of times a
service was used. Sum over all types of services and over rape, assault and stalking.

Total direct costs from intimate partner violence against adult women in the U.S. were over
4 billion dollars in 1995, including both mental health and medical care costs. Note that this
cost estimate does not include direct costs associated with the police and judicial system.

Source: CDC (2003).




                                                                                                 8

developing (and quite possibly developed countries as well), since GBV (and especially
intimate partner violence) may hinder women's participation in political and social
movements that would lobby for the provision of services (i.e., express the marginal
willingness to pay to politicians and policymakers). In addition, social norms may
inhibit women from acknowledging that violence has taken place and seeking help, in
essence artificially depressing their willingness to pay for (or at least use) services. Even
abstracting from these issues, direct cost estimates cannot be interpreted without a base of
reference such as GDP or cost estimates for other social issues.

 If estimates of direct costs of GBV are not particularly useful in a developing country
context, what options are left to document the socioeconomic costs of GBV? One option
is to concentrate on estimating indirect costs, as is discussed in the following subsection.

Indirect cost estimates: accounting, econometric and other approaches

Indirect cost estimates tend to highlight the impact of GBV on productivity and earnings
of women and ­ occasionally ­ of male perpetrators. Studies have focused on foregone
earnings due to death and lost productivity (CDC, 2003), job loss, lost productivity of the
women, lost productivity of the abuser due to incarceration, and mortality (Laurence and
Spalter-Roth, 1995), loss of tax revenues due to death and incarceration (Greaves et al.,
1995), and decrease in women's earnings (Morrison and Orlando, 1999; Sánchez et al.
2004; CDC, 2003).

Three approaches have been used to-date to produce indirect cost estimates. The first,
quite similar in spirit to the direct cost estimates above, is based on an accounting
approach. The total number of days of paid work or household chores lost due to GBV
(which is identified by responses to a survey question) is multiplied by the mean daily
earnings to yield a monetary estimate of lost earnings, whether this incapacitation is
temporary (due to injury) or permanent (due to death or incapacitating injury).11

In the case of recent estimates produced by CDC (2003), mean daily earnings are
calculated for the mean age of women affected by the various types of intimate partner
violence (rape, physical assault or stalking). In the case of non-paid household chores, an
imputed wage is used. Women in the U.S lost 10.1, 8.1 and 7.2 days of paid work,
respectively, from IPV stalking, rape and physical assault. The number of days of
household chores lost from these three types of IPV was even greater, at 12.7, 13.5 and
8.4 days. Multiplying these days lost by market or imputed wages yields a wage loss of
U.S. $1.7 billion in 1995; wages and productivity lost due to premature death alone were
U.S. $892 million (CDC, 2003).

One of the weaknesses of the accounting approach--whether used to estimate direct or
indirect costs--is that any list of categories is essentially arbitrary and alternative
categorizations can always be devised (Buvinic and Morrison, 1999). A more serious
weakness is that key categories of costs can be left out of the calculations. For example,

11Some estimates of the productivity impact include the reduction in earnings of male perpetrators due to
their incarceration (Laurence and Spalter-Roth, 1996).


                                                                                                         9

most direct cost estimates of the impact of intimate partner violence measure increased
use of health, judicial and social services by women, but do not include measures of the
impact that witnessing or being a victim of family violence may have on children.12

A second approach to calculating indirect costs of GBV has been to econometrically
estimate the impact of GBV on women's labor force participation and earnings. This
approach, employed by Lloyd (1999) and Farmer and Tiefenthaler (1999) for the U.S.
and Morrison and Orlando (1999) for Chile and Nicaragua, requires micro data sets that
contain standard labor force information on women's participation, hours worked and
earnings, as well as detailed information on women's experience with GBV. A reduced
form earnings equation for women is estimated, including a selectivity correction for
labor force participation. Using this approach, Lloyd and Morrison and Orlando all find
that intimate partner violence has no impact on women's labor force participation after
controlling for standard factors such as age, education, marital status, and number of
children in the household.13 Famer and Tiefenthaler, on the other hand, find that labor
force participation is higher for women who suffer abuse.

After estimating labor force participation equations, Morrison and Orlando (1999)
estimate earnings equations with standard explanatory variables and various variables
that measure the presence of intimate partner violence. The presence of any kind of
intimate partner violence--whether sexual, physical or psychological--is associated with
between a 34 and 46% reduction in earnings. Morrison and Orlando attempt to address
potential simultaneity between earnings and violence by using an instrumental variables
approach to instrument for violence; this econometric approach--while an innovative
method to allow estimation of the productivity impacts of intimate partner violence--has
several weaknesses.14

One approach that avoids the econometric complications of instrumental variables
estimation is propensity score matching. (See Box 2 and Section 4 for a description of
this methodology and an application to Colombia). Sanchez et al. (2004) recently used

12These impacts may include: poorer performance in school (Larraín et al., 1997); increased probability of
delinquency, both as a juvenile and as an adult (Windom, 1989; Dahlberg, 1998; Thornberry et al., 2001);
children leaving abusive homes to live on the street (Hernández Rosete, 1998); substance abuse (Molnar et
al., 2001); attempted suicide (Dube et al., 2001); and higher probability of committing family violence as
an adult (Strauss, et al 1980).
13This result is open to various interpretations. Lloyd suggests that two countervailing behavioral
responses to violence cancel one another out: some women who suffer abuse may be less likely to
participate in the labor market, but others may be induced to seek employment to acquire the resources
needed to negotiate a violence-free relationship or to exit the relationship altogether (Lloyd, 1999: 386).
Other candidates for countervailing factors include some women having lower labor force participation
rates because violence is used instrumentally by male aggressors to control women, while other women
who participate in the labor force trigger violence against them for this same control reason. This
simultaneity between violence and participation may offset the expected negative relationship between
violence and labor force participation. More research is needed, probably of a qualitative nature, to
disentangle the relationship between violence and labor force participation.
14The instruments for violence include physical abuse in the women's childhood and the number of times
per month the husband arrives home drunk. Both of these factors, however, may also have a direct impact
on earnings by affecting self-esteem and thus effort and ability. Also, the use of IV techniques in relatively
small samples can be problematic. (These observations were contributed by Vijayendra Rao.)


                                                                                                           10

propensity score matching to gauge the impact of intimate partner violence in Colombia
on a range of health and labor market outcomes. Based on a sample survey of over 2,000
women aged 15 to 49 in the cities of Bogota, Barrancabermeja and Barranquilla, Sanchez
et al. found that women who experienced moderate violence would have earned
approximately $60 more per month had they not been so victimized, and that women who
suffered severed violence would have earned more than $100 per month more--this
compared to mean monthly earnings for the entire sample of U.S. $142.



 Box 2. A non-parametric approach to estimating the wage impacts of intimate partner
 violence: the case of Colombia

 Step 1. Estimate probit equation of risk factors of intimate partner violence. Explanatory
 variables include: age, education of women and man, household wealth, consumption of
 alcohol, consumption of drugs, intimate partner violence in woman's origin household,
 proportion of women and children among household members, measure of household
 density (over-crowding), and labor force participation of man and woman.

 Step 2. Use propensity score matching--in this case 5-to-1 matching--to pair women who
 were victimized by violence to those who--although not victimized--have similar
 estimated a-priori probabilities of being victimized. The methodology requires that the
 range of a priori probabilities be similar for treatment and control groups, i.e., that there is
 an area of common support.

 Step 3. Compare means of matched "treatment" and "control" groups for the variables of
 interest. In this context, the treatment group is composed of women who suffered intimate
 partner violence; the control group is composed or their "matches" with similar a priori
 probability of suffering violence.

 Among the key results from this study were that women who were victims of intimate
 partner violence had:

     Significantly lower earnings (see text for description)
     Slightly lower rates of labor force participation (for some types of violence)
     A higher probability of suffering miscarriages or problems in childbirth
     Children more likely to have negative health outcomes such as recurrent coughing,
     severe respiratory infections, severe diarrhea, and high fever.
     Little or no impact on children's educational outcomes
     No impact on children's nutritional status


 Source: Sanchez et al., 2004




                                                                                                  11

Beyond direct and indirect costs: global measures of social costs associated with GBV

An option for estimating the socioeconomic costs associated with GBV--and one
frequently employed by economists to establish the market value of non-market goods--
is to estimate the willingness of individuals (and by extension society) to pay for lives
free of GBV, using contingent valuation methodology.          This methodology, to our
knowledge, has been used only once to estimate the willingness to pay for the reduction
of gender-based violence--and then to measure willingness to pay for violence
prevention services (see Sorenson, 2003).       One can only speculate as to why this
methodology has not been more frequently used. First, as the CEDAW and Belém do
Pará conventions make explicit, living lives free of violence is a fundamental human
right; estimating the willingness to pay for a fundamental right, while demonstrating the
importance society attaches to an issue, may in itself be controversial.        A second
unattractive feature of willingness to pay estimates is that they are sensitive to income
levels and income distribution.


    B. Methodologies for estimating the non-monetary impact of GBV

Several different methodologies have been employed to gauge the impacts of GBV,
without attempting to establish the monetary value of these impacts. They include:
estimating the disability-adjusted life years (DALYs) lost due to violence, "naïve"
comparisons of the mean values of outcome variables of interest between women who
have suffered violence and those who have not; and an emerging literature that uses
propensity score matching to improve the quality of the control group of women non-
victims before performing comparisons of means.


Disability-adjusted life years

The disability-adjusted life years (DALY) methodology has the great advantage of
including years lost due not only to premature mortality, but also due to disability or
illness. The first estimates of the DALY cost of gender-based violence were produced by
Heise et al. (1994), who estimated that more than nine million disability-adjusted years of
life are lost each year worldwide as a result of rape and family violence, more than that
from all types of cancer and more than twice that lost by women in motor vehicle
accidents.    More recently, Lozano (1999) estimated that rape and intimate partner
violence against women were the third most important cause of DALYs lost in Mexico
City--behind diabetes and perinatal conditions, but ahead of auto accidents, congenital
anomalies, rheumatoid and osteo-arthritis, cardiovascular disease, stroke and pneumonia.
(See Box 3 for more details on this study.)

DALY estimates produce a number--x number of years lost--but as can be seen from
the previous examples it is necessary to compare this number to DALY estimates for
other conditions in order to make some sense of the number. If this comparative exercise



                                                                                        12

is undertaken, DALY estimates are useful for dimensioning the importance of GBV
relative to other public health problems. The weakness of this approach is that outcomes
which do not result in mortality or morbidity, such as lost productivity, increased future
criminality or more street children, are not captured in the DALY estimates; moreover,
DALY calculations are methodologically complex and data-intensive.



Box 3. Estimating the disability-adjusted life year cost of violence against women:
evidence from Mexico City

Step 1. Estimate the DALYs due to premature death due to violence against women.

            Obtain number of female homicide victims in Mexico City.
            Based on police statistics on identity of killer, establish the percentage of
            homicides which were committed by male partners (whether common law or
            formally married). In Mexico City, 28% of killers were male spouses, 30% were
            relatives or known to the victim, and 42% had unknown identities.
            Based on the age profile of victims and data on life expectancy, establish the total
            number of life years lost due to premature death at the hands of a male partner.
            Undertake a similar calculation for deaths of women attributable to sexual
            violence by non-partners, and for suicide victims whose suicide was a result of
            intimate partner violence or rape.

Step 2. Establish the DALY due to years lived with disability

            Estimate DALY consequences of physical injuries. The study documented that
            50% of women seen in emergency rooms were presenting injuries that resulted
            from intimate partner violence. Also include those seen due to injuries associated
            with rape or suicide (if related to rape or IPV).
            Estimate DALY due to the psychological secuelae of IPV and rape, as well as the
            impact on reproductive health.

Step 3. Sum the DALYs due to premature death (42% of the total) and those due to disability
(57% of the total) to get the global DALY figure due to violence against women. Another
breakdown of the DALYs lost is 53% due to physical secualae (death, physical injury, etc.)
and 47% due to non-physical secuelae. Total disability-adjusted life years lost for 1995 were
27,200--the third most important source of DALYs lost for women in Mexico City.


Source: Lozano, 1999.




Naïve comparisons of means

In both developed and developing countries, there is a growing literature that documents
the negative health outcomes associated with GBV. Studies have examined the impact on
women's well-being, including:                reproductive health (unwanted pregnancies,



                                                                                                13

contraceptive use high-risk sexual behavior, exposure to HIV/AIDS, high-risk
pregnancies, and other gynecological problems);                    physical well-being (injuries and
deaths); mental health (depression, post-traumatic stress disorder, suicide, and alcohol
and drug abuse); and participation in community activities. Impacts on children's mental
and physical health, as well as their educational performance, have also been documented
(Heise, Ellsberg and Gottemoeller,1999).

The most methodologically rigorous of these studies compare abused women against a
control group of women who did not suffer GBV, and use standard statistical tests to
identify statistically significant differences in health outcomes (Campbell, 2002a;
Campbell et al., 2002b; Coker et al., 2002, 2000a and 2000b; Jewkes, 2000; Kishor and
Johnson, 2004; Rao, 1998). A few studies (such as Ellsberg et al.,2001) combine this
comparison with in-depth interviews with women to aid in the interpretation of the
quantitative data.

The conclusions of these studies are only as strong as the control groups are rigorously
selected. At the lax end of rigor are studies are based on random surveys of women in a
certain geographical area and simply compare mean values of outcome variables between
women who reveal having been victimized by violence and those who respond that they
were not victimized. The obvious concern is that women who respond that they were not
victimized are not really similar to those who were victimized, either in terms of
observable variables (such as age, marital status, personal history of witnessing violence
when young, or other well-established risk factors for violence). More rigorous are
studies which match women on the basis of observable variables, thus ensuring rough
comparability between victimized women and the control group--although these studies
are still open to the criticism that unobservable factors account for the differences
observed between the two groups of women.


Propensity score matching

Propensity score matching (PSM) deals with the limitations of naïve comparisons of
means by constructing treatment and control groups not on the basis of values of
observable variables, but rather on the estimated probability of suffering intimate partner
violence.15

PSM offers several benefits vis-à-vis other methodologies. First, if the impact of intimate
partner violence on health and labor force outcomes are heterogeneous across individuals,
matching may produce very different results than methodologies which do not drop
observations for which there are not common values in both control and "treatment"
groups (Smith, 2000). In PSM, this is known as the "common support condition", and it
requires that the range of estimated propensity scores be similar in treatment and control


15PSM was initially used almost exclusively for program evaluation, and the terminology that has grown
up around the technique reflects this. Thus, treatment groups are participants in the program, and the
control group is of non-participants. In the case of intimate partner violence, the control group is of women
who did no suffer intimate partner violence, while the "treatment" group is of women who did.


                                                                                                           14

groups. Heckman et al. (1999) have shown that failure to compare the treatment and
control groups within this region of common support is the single most important source
of bias--one that overshadows the "classic econometric problem of selection bias due to
differences in unobservables (Jalan and Ravallion, 2003: 21)."

Second, as compared to regression analyses (see the section on indirect costing
methodologies for a summary of regression analyses that have been used to gauge the
impacts of intimate partner violence), PSM is non-parametric. Consequently, it avoids
the sensitivity to functional form choice that can plague regression models.16 Some
evidence suggests that avoiding functional form restrictions can significantly reduce bias
(Dehejia and Whaba, 1998; Smith and Todd, 2000; Smith, 2000).

Of course, PSM has its own set of econometric complications, including the need for
quite large sample sizes, not offering advances over traditional regression analysis in the
treatment of unobserved heterogeneity, and not explicitly dealing with issues of
simultaneity between earnings and violence. In addition, the quality of PSM estimates
depends on conditional independence, namely that the error term in the probability model
used to estimate the propensity score is uncorrelated with the outcomes of interest. PSM
is also very sensitive to omitted variable bias.

We use PSM in the next section of this paper to estimate the impact of GBV on women's
health outcomes, use of health services and labor force participation, as well as the
impact on children's health outcomes and educational performance.


4. Estimating the Impacts of GBV on Women and their Children: An
    Approximation using Propensity Score Matching



A. Prevalence of GBV in Peru, Haiti, and Zambia


The data

We use data from National Demographic and Health Surveys (DHS) which collect
information at the individual and household level. The main respondents are women aged
15 to 49 years. These surveys typically have large sample sizes (between 5,000 and
30,000) and take place periodically every 5 years. Some of the aspects covered regularly
in a DHS are: fertility, family planning, reproductive health, young children's health,
nutritional status, sex of each household member, age, education, marital status, area of
residence, and a variety of household characteristics. The data is collected by National


16While the use of PSM relieves the researcher from needing to make difficult choices about functional
form, it does not relieve him/her from deciding which set of variables to include in the logit or probit
specification. Heckman et al. (1999) document that the estimates that emerge from the matching procedure
can be quite sensitive to the specific vector of explanatory variables included (Smith, 2000).


                                                                                                         15

Statistics Institutes in collaboration with the Measure DHS+ program developed and
implemented by ORC Macro (Macro International Incorporated)                         17. ORC Macro also
creates codes and processes data files to guarantee consistency and comparability across
countries.

Recently, the Measure DHS+ program has begun to include a domestic violence module
in some countries. The objective of this initiative is to study the linkages between
domestic violence, health , and demographic outcomes.18 We selected Peru, Haiti, and
Zambia from a group of 11 countries that have data on domestic violence. The choice of
these countries was driven in part by a desire to choose countries in which the health
impacts of intimate partner violence have not been examined, and in part by a desire to
have representation of both Latin American and African countries.19

In Peru and Haiti a domestic violence module was added to the questionnaire of the 2000
DHS. In Zambia, the DHS module was added to the questionnaire for the 2002 DHS. In
Haiti and Zambia a randomly selected woman in the household answered the domestic
violence module. In Peru, all women 15-49 in the household answered the domestic
violence module. This explains, in part, the large sample size of the data set
corresponding to Peru.

The most accepted method of measurement of GBV in a survey is the Conflict Tactic
Scale (CTS).20 In the Measure DHS+ program some countries decided to use a modified
CTS, while some countries measure GBV by first asking whether the respondent has
experienced violence and proceed with more detailed questions if the answer is
affirmative.21 The Haiti DHS 2000 used a modified CTS, while the Peru (2000) and
Zambia (2002) surveys used initial questions to determine victimization.


17This program is part of United States Agency for International Development's (USAID) initiatives to
help developing countries collect and use data to monitor and evaluate population, health, and nutrition
programs. For more information on DHS and available data sets see : www.measuredhs.com

18See Kishor and Johnson (2004) Chapter 1 for a detailed description of the domestic violence module in
DHS. The chapter includes a discussion of sampling methods, ways to ensure the ethical collection of
violence data, and attempts to minimize the underreporting of violence.
19Three DHS countries were not analyzed--Nicaragua, India and Colombia--because of the existence of
high-quality, existing work on the health impacts of gender-based violence. For Nicaragua, see Ellsberg et
al. (2001). For India, see Duvvury et al. (2000) and Rao (2002). For Colombia, Sanchez et al. (2004) have
produced the only other study that to our knowledge uses PSM to examine the health impacts of intimate
partner violence.
20The CTS method consists of a set of questions about specific acts of violence, ranging from mild
psychological abuse to life -threatening acts of violence. This method reduces the self censoring that occurs
when a violence victim is asked directly if she has been hurt by her partner. It also allows to collect data on
"mild" acts of violence for some women who may choose not to answer questions regarding brutal acts of
physical violence or sexual violence by their partners or relatives. According to Kishor and Johnson
(2004), the CTS avoids report inconsistencies due to subjective and cultural perceptions about what
constitutes violence.
21See Kishor and Johnson (2004) for details on measurement of GBV in the Measure DHS+.
Questionnaires and complete descriptions of the DHS including the Violence Module are available at
www.measuredhs.com




                                                                                                             16

The type of GBV variables available vary from country to country. While the violence
questions in DHS generally focus on domestic violence by partners and relatives, the
Haiti (2000) and Zambia (2000) surveys include some questions about violence
perpetrated by strangers, professors, and employers. Sexual violence was not included in
the Peru (2000) DHS. Similarly, the Peru (2000) DHS did not include questions about the
timing of violence episodes.

Even though all three countries included some variables regarding                           psychological
violence or emotional violence, we decided not to use them as they varied significantly
across countries. Except for the Haiti (2000) survey, which used a modified CTS, the
proxy variables for psychological violence were insufficient to characterize the
phenomenon.22

In all countries, we select a sub-sample of women who are currently married or living
with a partner. This is consistent with the vast majority of work on GBV in developing
countries. This population group is of particular public policy interest because they face
a higher risk of intimate partner violence than women who do not live with a partner. In
addition, this group is more homogeneous than the whole sample in terms of how GBV is
experienced, age range, fertility, and other risk factors; thus, facilitating the
implementation of PSM estimation. Unfortunately, the selection of women who are
currently living with a partner implies that our study does not include women who
currently do not live with a partner but were living with a violent partner in the past and
were able to leave him.23

Using DHS data has two main advantages. First and foremost is the availability of
reasonably comparable data across countries not only on women´s experiences with
gender-based violence, but also on a series of outcomes of interest such as health status of
women and their children, and some basic data on labor force participation. A second
advantage of DHS data is that they have been cleaned and made publicly available.

DHS data do have some disadvantages, though. Research in Nicaragua and Colombia
has documented that DHS tends to underestimate gender-based violence, when compared
to stand-alone violence surveys such as the Conflict Tactics Scale or WHO´s multi-
country survey on gender-based violence. Ellsberg et al. (2001) find that DHS data leads
to prevalence estimates in Nicaragua which are about half those generated by specialized
violence surveys (28% versus 52%). Second, DHS surveys do not universally distinguish
between current and lifetime violence (some surveys do, but others do not), and DHS
modules contain little information about the aggressor and triggers of violence (such as
employment status of the aggressor).

22In the case of Peru, the questions on psychological violence by partner included yelling, limiting contact
with friends and family, attacking the woman's self esteem, and threats of abandoning the house. In the
case of Zambia, there were no questions on psychological violence in the survey.
23This particular group of women could provide invaluable information regarding factors that allow
women to escape a violent situation. Despite significant obstacles, many women do eventually leave
violent partners, although this departure may take many years. A study for Leon, Nicaragua , found that
70% of abused women eventually left an abusive partner (Heise et al., 1999).


                                                                                                          17

Prevalence of GBV for currently married or co-habitating women

Table 4 summarizes indicators on prevalence of GBV for currently married (or co-
habitating) women in Peru, Haiti, and Zambia.24 In Zambia, almost 42% of women had
experienced physical violence at the hands of an intimate partner sometime in their lives;
in Peru, this figure was almost 40%, while in Haiti less than 15% of women had been so
victimized. Data on current violence (occurring in the 12 months prior to the survey)
inflicted by intimate partners are available only for Haiti and Zambia, where 12.5% and
22% of women were affected, respectively.

Lifetime sexual violence at the hands of an intimate partner is significantly more
common in Haiti than in Zambia--18.9% versus 5.5%.                           17.4% of Haitian women
reported current sexual violence by intimate partners. In Zambia--as in the case of
physical violence by intimate partners--the sample size drops significantly for current
sexual violence. 6.6 % of married or cohabiting women report being affected by this type
of violence.

Physical violence by relatives or other members of the household is high in Peru and
Zambia (26.6 and 22.0%, respectively); in Haiti, the corresponding figure is 2.9%.
Physical violence by strangers (available only for Haiti) is quite uncommon. Finally,
sexual violence by a male other than the intimate partner (available only for Zambia)
affected 13% of women at some point of their lives.


B. Logit estimation of risk factors for GBV

GBV is a complex phenomenon that is shaped by forces that operate at different levels.25
An ecological framework that includes factors which operate at the individual,
relationship, community and societal levels consequently is an appropriate model with
which to examine the risk factors for GBV.26

The most appropriate way to identify risk factors for GBV is in a multivariate framework
rather than through bivariate correlations. A multivariate framework allows one to
control for the presence of potentially confounding factors and to establish the relative
importance of different factors. Most analyses of risk factors have been undertaken in the
United States and Europe, although there are some recent findings from South Africa,



24For prevalence data for the overall DHS sample see Kishor and Johnson (2004) or www.measuredhs.com
for country reports.
25The first three paragraphs of this section are based upon Morrison, Ellsberg and Bott (2004).
26The term "risk factors" is used instead of the more common term "determinants". The term determinants
implies a mechanistic relationship between variables: if a man abuses alcohol, for example, intimate partner
violence will result. This is clearly not the case; alcohol abuse increases the likelihood of intimate partner
violence, but does not mechanistically indicate the presence of violence (Morrison, Ellsberg and Bott,
2004).


                                                                                                           18

Bangladesh, Uganda, Chile and Nicaragua (Jewkes, 2002; Koenig et al. 2003, Larrain,
1994; Ellsberg et al. 1999).

It is also important to build econometric models that are specific to the different
manifestations of GBV. As can be seen below, while some of the risk factors that emerge
as statistically significant in the logit regressions are common across the different types
of violence, there is much variation in the significance of particular variables across
different types of violence..

Table 5, taken from Morrison, Ellsberg and Bott (2004), identifies risk factors for
intimate partner violence that have been found to be significant in recent empirical
studies.27 Unfortunately, not all the variables identified as potentially important are
available in the DHS surveys; we identify with italics those variables which are
available--albeit imperfectly specified in some cases--in at least one of our countries of
interest. Two additional variables, household size and structure (nuclear versus non-
nuclear) and urban versus rural residence are also included in our logit specification
despite not being present in Table 5.28

We estimate a separate logit model for each of the different types of GBV we will
examine: physical violence by intimate partner; physical violence by relative; and sexual
violence by intimate partner.29 For each type of violence, we estimate the determinants
of current violence (experienced in the 12 months prior to the survey) and lifetime
violence.

In general, the equations for physical violence against women by intimate partners show
a reasonably good fit. The equations for sexual violence by intimate partners have lower
overall significance than those predicting physical violence. As expected, the relevance
of each risk factor varies greatly by type of violence and whether the violence has
occurred in the past or is currently occurring (in the last 12 months).


Lifetime physical violence by intimate partners

Table 6 contains the logit estimation results for risk factors for physical violence that
occurred at any point in the lifetime. Age of the women has the expected negative sign


27Factors that increase the probability of experiencing intimate partner violence have a positive sign next to
them and factors that diminish the probability of experiencing violence have a negative sign next to them.
Some factors can have a positive or negative sign depending on the specific context.
28Our logit model is similar to the logit model estimated by Kishor and Johnson, who also use DHS data.
They, however, do not use these logit results as inputs to propensity score matching.
29The DHS survey also contains information on sexual violence suffered at the hands of relatives and
physical and sexual violence suffered at the hands of strangers for some of the countries. Logit equations
for these types of violence are not reported for the following reasons: i) the sample of women who
answered questions on sexual violence by relatives was much smaller than the original sample size; ii) the
ecological model of risk factors of domestic violence was not relevant to predict physical or sexual
violence by strangers; iii) DHS data sets do not contain variables that are risk factors for these types of
violence.


                                                                                                           19

only in Haiti for women over 45. Surprisingly, in Peru older women experience a higher
risk of violence--which may be due to cohort effects, i.e., a generational transition
towards lower physical violence levels. In the case of Zambia, age is never statistically
significant. Women's age at marriage (or first intercourse age) is significant in all
countries and an older age at marriage uniformly reduces the probability of experiencing
violence. Women who have been in more than one union have a lower probability of
having suffered physical violence in their lifetime in Peru and Zambia, but the variable is
not significant in Haiti. The protective effect of having had more than one union may be
illusory; it may in fact be picking up the fact that women left a relationship where
physical violence was imminent.

In Peru, the risk of violence increases monotonically with the number of children, but in
Zambia women with one or two children have higher risk, while those with more children
have the same risk level as those without children; this result is difficult to interpret. In
Peru and Haiti women who were victims of physical violence as children have a
significantly higher probability of experiencing intimate partner violence as adults;
unfortunately, this variable was not available for Zambia. Alcohol abuse by the partner
significantly raises the probability of violence in the two countries--Peru and Haiti--
where this variable is available; in Peru, it increases the probability of violence more than
nine times.

Having an older male partner is generally associated with a lower probability of violence.
In Peru and Zambia this effect is generally monotonic, but in Haiti violence does not
decrease until a woman is older than 45. In Peru, the male partner having more education
than the female partner is associated with lower violence; in Haiti, the opposite effect
prevails. Marital duration is significant only in Zambia, but it has an unexpected sign:
longer marriages are associated with more violence, not less. Control by the husband
increases the probability of violence, while participation of the woman in economic
decision making in the household reduces it. In Zambia, women living in nuclear
households have a higher probability of suffering of physical violence which may be due
to higher isolation of the woman from her relatives and the relatives of her partner.
Wealth is statistically significant only in Peru, where it has a non-linear impact on
violence: households in the third quintile register the highest probability of violence,
followed by second quintile households.30 Fourth and first quintile households have equal
probabilities, and the richest quintile registers the lowest probability of physical violence
by a male partner.

Finally, at the community level, urban households register a higher probability of
violence than rural households in all three countries.

The model is globally significant at the one percent level in all countries. It correctly
predicts between 59 percent (Zambia) and 84 percent (Haiti) of all cases of physical
violence by partner. The model improves the forecast of physical violence by male


30 We use wealth to proxy for economic hardship (see Table 5). The wealth index, in the DHS data set, is
based on house conditions, location, access to water and sanitation, and availability of certain household
goods (bicycle, refrigerator, etc.).


                                                                                                          20

partner over the predictions of a "naïve" model--based simply on the frequency of
violence--in the three countries, but the improvement over the naïve model is quite
modest in Haiti and Zambia.


Current physical violence by intimate partners

Table 7 shows the result of the estimation of a logit model predicting the probability of
current (last 12 months) physical violence by a male intimate partner. Data on the timing
of violence are not available in the case of Peru; thus, results are reported only for Haiti
and Zambia. The most relevant risk factors behind current physical violence are: age of
the woman (older women facing lower probabilities of experiencing current violence),
number of unions, experiencing violence as a child, male partner's alcohol abuse, marital
duration (in Haiti longer duration increases the risk of violence, while in Zambia it
reduces the risk of violence), family structure and area of residence (urban areas face a
higher probability of violence).

For Haiti, the results for current violence are very similar to those for lifetime violence; in
the case of Zambia, the results look quite different.31 The model is globally significant at
one percent. It correctly predicts more than 65 percent of all cases of current physical
violence by partner. The model improves on the forecast of a "naïve" model in both
countries--quite significantly in the case of Zambia and modestly in the case of Haiti.


Current and lifetime sexual violence by intimate partners

Tables 8 and 9 summarize the results from estimating a logit model to predict lifetime
and current sexual violence by partner in Haiti and Zambia; data on sexual violence are
not available for Peru. The risk of ever experiencing this type of violence diminishes
considerably with age for women in Haiti. In Zambia, chronological age is not significant
but women who were older at the moment of first sexual intercourse have a lower
probability of experiencing sexual violence by a partner. Women with children also
experience less risk of sexual violence in Zambia. Completing primary education reduces
the risk of experiencing sexual violence for women in Zambia. Alcohol consumption by
the partner increases the risk of sexual violence in Haiti, but only up to a certain extent
(perhaps due to the fact that alcohol abuse may have a negative impact on sexual
performance in males). Large age differences between partners reduce the probability of
sexual violence when the male partner is older than the female (in Haiti). Control by the
male partner more than doubles the odds of experiencing this type of violence for women



31Age of the woman is statistically significant for current violence, but not for lifetime violence. Women's
age at marriage matters for lifetime violence, but not for current violence. Number of unions and duration
of marriage are significant for both lifetime and current violence, but the direction of the impact differs.
These results are driven by the fact that the question on current violence in the Zambia DHS was not well-
worded, and many women did not answer the question. Thus, the sample size for the current violence logit
is less than half as large as in the case of lifetime violence.


                                                                                                            21

in Haiti. Women in urban and middle income households face higher risks of sexual
violence by partner than their counterparts in rural areas or poor or wealthy quintiles.

In spite of being globally significant at 1 percent in Haiti and 10 percent in Zambia,
the models of sexual violence are not very impressive. In the case of lifetime violence,
they offer little improvement in predictive power over the naïve model in Haiti, and
perform marginally worse than the naïve model in Zambia.

The risk of experiencing sexual violence by partner in the last 12 months (see Table 9)
could only be estimated in the case of Haiti due to the large number of missing values in
the dependent variable for Zambia. Results are similar to the results obtained in the
model predicting the probability of ever experiencing sexual violence by a male partner.

Lifetime physical violence by relatives

Table 10 summarizes the results of the logit model predicting physical violence inflicted
on women by relatives or other household members (aside from their male partner). The
risk of experiencing this type of domestic violence diminishes with age in the three
countries. Women who are in their second ­or higher- union have a higher risk of
experiencing violence by relatives (perhaps the relatives of her new partner). The odds
are reduced significantly for women with children--especially large numbers of
children--in Haiti and Zambia. In Zambia, women who have secondary education
paradoxically face a higher probability of physical violence from their relatives than less
educated women. In Peru, alcohol abuse by the male partner increases physical violence
of relatives towards his female partner. This form of violence is more likely in urban
households and, in the case of Peru, in less wealthy households.

This model is globally significant for Peru and Zambia, but not for Haiti. Even though the
model correctly predicts a large percentage of cases, it does not improve on a naïve
model in any of the three countries.


C. Propensity score matching results: Effects of violence on health , education, and
   employment outcomes

Using the estimated probabilities from the logit models above, we estimate propensity
scores for each type of violence. The propensity scores are used, in turn, to construct a
"propensity" matched control group: women who have not experienced violence but have
a similar a priori probability of experiencing violence and similar values of the risk
factors as victims (i.e., both the common support condition and balancing property are
met). We employ one-to-one matching in which each woman who suffered violence is
matched to one woman in the control group with the closest predicted value from the
relevant logit equation.32Once the groups are constructed we estimate average treatment



32We employ the Nearest Neighbor Matching algorithm for STATA (pscore), recently created by Becker
and Ichino, to estimate average treatment effects.


                                                                                                22

effects of the occurrence of violence. The significance of differences between the two
groups is tested using bootstrapped standard errors.

The tables presenting the results of the PSM that follow are organized into six panels that
summarize the effect of each type of GBV on: i) women's health; ii) women's use of
health services; iii) women's employment; iv) health of children aged 0-5; v) educational
achievement of children older than 7; and vi) mother's use of violence to discipline
children. It is crucially important to separate women's health outcomes from women's
use of health services, since the impact of GBV on women's health outcomes is
unambiguously negative, while its impact on the use of health services is much less so.
The use of health services and victimization by violence in fact may be positively
correlated, since violence may compel women to seek medical care, at which time other
health services are provided. In all the tables, the average treatment effects are presented
as the mean value of women who suffered GBV (or the mean value for their children)
minus the mean value of the control group of women unaffected by GBV (or the mean
for their children). Table 11 contains descriptive statistics for each outcome variable.

Lifetime physical violence by intimate partners

Table 12 reports the difference in outcomes of interest for women (and their children)
who have experienced physical violence at the hands of intimate partners and a control
group of women unaffected by physical violence by their partners.

In general, the health outcomes of women who suffered physical violence are markedly
worse than that of their matched peers. They are far more likely to have:

         complications during delivery: in Peru, the percentage of women who had
         complications was 7 percentage points higher among women victims of physical
         violence.
         terminated their pregnancy before term: in Peru, women victims of physical
         violence on average terminated more pregnancies than non-victims.
         contracted sexually transmitted diseases: in Peru and Zambia, the percentage of
         women who contracted STDs was 4.0 and 1.9 percentage points higher
         (respectively) among women victims of physical violence.
         suffer from genital sores or ulcers: in Haiti and Zambia, the percentage of women
         who contracted genital sores or ulcers was 4.3 and 2.8 percentage points higher
         (respectively) among women victims of physical violence
         anemia: in Haiti, women who suffered physical violence had on average, more
         severe anemia than non-victims. In Peru, however, women who suffered
         physical violence had ,on average, less severe anemia.

In fact, the only health outcome measure for which there is not evidence of the pernicious
impact of physical violence is women's nutritional status, as measured by weight for
height. On this measure, there is no statistically significant difference between victims
and non-victims.




                                                                                          23

The usage of health services is another issue. In general, women affected by physical
violence tend to use health services more intensively, although the evidence for this
assertion is not overwhelming. In one out of the three countries, women victims of
physical violence were more likely to visit a health facility, have births assisted by a
health care professional, and use contraceptives.33 On the other hand, in Haiti, women
victims were significantly less likely to receive prenatal care.

Total fertility shows no clear relationship to physical violence. Nor are family planning
needs or outcomes clearly linked to physical violence: in Peru, women victims are less
likely to have unmet family planning needs and more likely to describe their last child as
"wanted", but in Zambia victims are more likely to have unmet family planning needs.

There is strong evidence that women affected by physical violence are more likely to be
in paid employment (ranging from 2.2% more likely in Haiti to 6.0 and 6.2% more likely
in Zambia and Peru, respectively.

With relation to children's health outcomes, children of women who suffer physical
violence at the hands of intimate partners are more likely to suffer diarrhea (6.7
percentage points more likely in Peru), suffer anemia (19.2 percentage points more likely
in Haiti), and trail children of unaffected women in height for age (Peru). On the other
hand, children of women victims are more likely to be immunized in Peru. Other child
health outcomes such as under-five mortality and chronic coughing show no relation to
physical violence against the mother.

In general, children's educational outcomes are unaffected by physical violence against
their mothers, with the exception of Peru where children of women victims are more
likely to attend school and less likely to be behind in school.

Finally, there is evidence from Peru that women who suffer physical violence are more
likely to use violence to discipline their children. This leads to an inter-generational
transmission of violence, since children who are victims of violence are likely to
reproduce violence later in their lives.34


Current physical violence by intimate partners

Table 13 presents the limited evidence on the impact of current physical violence by
intimate partners, based on results from Haiti.35 The negative impact on women's health



33As suggested above, this may simply be the result of women seeking treatment for injuries sustained
because of physical violence by intimate partners, and then being referred to other health services. In the
other two countries, the differences between the two groups were not statistically significant.
34Note that there is scientific evidence also documenting the inter-generational transmission of violence if
children merely witness family violence--even if they are not direct victims.
35The evidence is limited because the Peru DHS survey does not allow identification of the time period in
which violence occurred. The Zambia question was badly worded, leading to a loss of more than half the
sample because of missing values. In addition, when we attempted to implement PSM with the smaller


                                                                                                           24

outcomes is supported by the fact that women in the victims group have more terminated
pregnancies and genital sores and ulcers. They are significantly more likely (17.6
percentage points) to have visited a health facility.

There is no measurable impact on women's labor force participation, and the effects on
children's health outcomes are modest: the only statistically significant difference is a
slightly higher rate of anemia among children of women victims.

The impact on children's educational attainment is large and statistically significant;
children of women victims, on average, have accumulated approximately one-half a year
less of education than school-aged children of women in the control group.




Lifetime and current sexual violence by intimate partners

Tables 14 and 15 report the impacts of lifetime and current sexual violence by intimate
partners in Haiti.36 The impact on women's reproductive health is similar to that of
current physical violence: women victims are more likely to have terminated pregnancies
(lifetime sexual violence), to have a sexually transmitted disease (current sexual violence)
and to have genital sores and ulcers (current and lifetime sexual violence). Women
victims seem to be better nourished and experience obesity more frequently, as measured
by weight for height. As is the case for current physical violence in Haiti, victims of
lifetime sexual violence are more likely to have visited a health facility. Impacts on
children--whether measured by health or educational outcomes--are not significant for
either current or lifetime sexual violence.


Lifetime physical violence by relatives

Table 16 reports results for this type of violence for Peru.37 In spite of having a higher
contraceptive use, women victims register a higher number of terminated pregnancies,
unwanted children, and delivery complications--the latter perhaps because these women
are also less likely to give birth assisted by a health care professional. Women victims
have higher weight for height than women in the control group. Finally, women victims
have a higher probability of working and earning cash than women in the control group.



Zambian sample, most of the variables of the logit model violated the "common support condition"
required by PSM. Thus, we do not report PSM results for Zambia.
36In the case of Peru, data on sexual violence are not available. In the case of Zambia, the results of the
logit model for lifetime violence did not predict violence well enough to constitute the base of the PSM. In
the case of current sexual violence in Zambia, missing values reduced the sample size precipitously,
preventing the estimation of a propensity score.

37 In the case of Zambia, the base logit model was not robust enough for the PSM estimation. In the case
of Haiti, none of the effects were statistically significant and consequently are not reported in Table 15.




                                                                                                           25

Child health outcomes are noticeably poorer for women victims, as measured by height-
for-age, weight-for-height and suffering from diarrhea. At the same time, educational
outcomes are stronger for children of women victims. One wonders if women may
remain in violent situations in exchange for economic support that allows their children
to attend school--as has been documented in the case of sexual violence against women
and girls in the African context (Leach, 2003). Finally, women who are victims of
physical violence by relatives are more likely to use violence to discipline their children
than women in the control group.


Overview of results

In general, we find strong impacts of GBV on women's health outcomes, especially on
reproductive health; GBV also increases--not surprisingly--women's usage of health
services.38 The fact that women who are victims of GBV have a more intensive use of
health services points out the high costs to society of GBV in terms of resource use.39

The impacts of GBV on paid work are strong: lifetime physical violence by intimate
partner increases the probability of paid work in all three countries. This result, of course,
may be the result of simultaneity between earnings and violence which is not resolved by
PSM. On the one hand, women who are victims of GBV may have lower levels of
productivity on the job and earn less. On the other hand, women who work outside the
home threaten male supremacy and hence may be more frequent victims of violence.40

The study finds strong evidence that GBV may lead to the inter-generational transmission
of violence.41 Women who are victimized by physical violence--whether at the hands of
intimate partners or relatives--are more likely to use violence in disciplining their
children; much scientific evidence has documented that children exposed to such
violence are more likely to engage in violent and delinquent behaviors.

The study also finds an impact of GBV on children's health outcomes, with the specific
manifestations varying by type of GBV and by country. Where impacts are found on
children's educational performance, they are of an unexpected direction, with children of
women victims more likely to attend school and less likely to be behind in age/grade



38Note that this is after controlling for urban or rural residence, household wealth and other variables which
would be present in a demand function for health services, so we are able to argue strongly that this finding
is attributable to GBV.
39Of course, this cost will and should increase as more women make use of services and as services to
survivors are expanded. Service use can be interpreted as a positive development: women are cognizant of
available services and willing to seek help.
40Note that the assumption made in estimating the logit equation for GBV was that labor force
participation was not a determinant of GBV; this assumption was necessary in order to use PSM to examine
the impact of GBV on labor force participation. The result of the PSM exercise--that women who suffer
GBV are more likely to engage in paid work--calls the validity of this assumption into question.
41Evidence on this link is particularly strong given the fact that it is less affected by reverse causality
issues.


                                                                                                            26

progression.42A possible explanation for this result is that women who suffer violence at
home try to keep their children outside of the home for as long as possible to protect them
from abuse.


5. Conclusions

This paper has described the various methodologies available for estimating the impact of
gender-based violence, whether a monetary value is attached to these impacts or not. It is
clear that the field has grown enormously since the first direct cost estimates were
produced more than ten years ago. It is also clear that these estimates have had a key role
in the public policy debate about gender-based violence.

Not surprisingly, all the methodologies in this paper have advantages and disadvantages
(as summarized in Table 3). Accounting approaches to estimating the costs of GBV are
straightforward to carry out, but may seriously underestimate the true social costs of
violence and may be especially problematic in a developing country context. Indirect cost
estimates of the impact of GBV on women's labor force participation and earnings have
been successful in highlighting the productivity impacts of gender-based violence and
have helped establish that GBV is an important to both economic development and
gender equality; at the same time, indirect cost estimates have suffered from econometric
weaknesses that have yet to be definitively resolved.

DALY measures of the health impacts of GBV have proved indispensable for firmly
placing the issue of GBV on the public health agenda, yet they have not been particularly
useful in terms of formulating policy responses to violence and seem to have had a
somewhat limited impact beyond the health sector. Naïve comparison of mean outcomes
between women who have suffered violence and those who have not have contributed to
identifying the specific health impacts on women and their children, but have been
criticized for being less than rigorous in their formation of control groups of non-victims.
Finally, propensity score matching adds rigor in the specification of control groups, but
does not solve simultaneity issues.

In sum, there is no perfect methodology with which to gauge the costs and impacts of
GBV. All methodologies have strengths and weaknesses, and the challenge is to choose


42It is interesting to compare our results using PSM to those obtained by Kishor and Johnson (2004), who
also used DHS surveys from these countries but who simply compared mean values for victims and non-
victims. Of course, the results are not strictly comparable, since Kishor and Johnson did not distinguish
between different types of violence and since their sample sizes are larger (we include in our sample only
currently married women and loose observations due to missing values in some variables in the logit
regressions). Kishor and Johnson found effects of GBV on number of children in Peru and Zambia which
are not present in our PSM results. (Interestingly, Heise, Ellsberg and Gottemoeller (1999), citing evidence
from Nicaragua, argue that the causality goes in the other direction: women who are victims of violence
are less able to control their fertility, and consequently have more children.) Kishor and Johnson also find
that GBV increases under-five mortality in all three countries, while we find no such effect. There are
other, more nuanced differences between the impacts detected by the two studies.




                                                                                                           27

the appropriate methodology given both data constraints and the potential consumers of
the estimates.

Future research on the costs and impacts of GBV has many promising fronts to explore.
First, it would be useful to employ the PSM methodology on richer data sets that include
more information on the women's labor force participation and earnings, as well as
household consumption patterns. This would allow the indirect costs of GBV to be
explored in a methodologically rigorous way. Second, more work needs to be done on
the relationship between income, wealth and violence. Any research has to deal with
simultaneity issues, which PSM does not resolve. Third, the results from this paper
suggest that GBV is more prevalent in urban areas--which is exactly opposite the
conclusion emerging from the WHO's multi-country study. Studies on this issue will
need to address under-reporting of violence, which is frequently assumed--without
evidence--to be a larger problem in rural areas; these studies necessarily will be
qualitative in nature. Finally, there is a need for more studies on the economic costs of
regionally-specific types of GBV such as female genital mutilation, dowry violence and
incest, in order to position this issue as what it is: not only a woman's issue, a public
health issue and a human rights issue, but also as an issue affecting poverty, development
and economic growth.




                                                                                          28

Figure 1: Percentage of women who reported physical violence by an intimate
partner in past 12 months and over lifetime

  Bangladesh (Urban)                                                        40%
                                              19%

Bangladesh (Province)
                                         16%                                  42%

        Brazil (Urban)                                    27%
                                 8%

     Brazil (Province)                                             34%
                                     13%

   Ethiopia (Province)                                                                  49%
                                                             29%

        Japan (Urban)                13%
                          3%

     Namibia (Urban)                                           31%
                                         16%

         Peru (Urban)                                                                    50%
                                           17%

      Peru (Province)                                                                                     62%
                                                       25%

     Thailand (Urban)                              23%
                                8%

   Thailand (Province)                                             34%
                                      13%

    Tanzania (Urban)                                             33%
                                        15%

  Tanzania (Province)                                                               47%
                                             19%

                Serbia                             23%
                          3%                                                            physical violence ever
                                                                                        physical violence past 12 months
               Samoa                                                         41%
                                            18%


Source: Unpublished data from the WHO Multi-country study on Women's Health and Domestic Violence
Against Women. The final published comparative report is forthcoming. Cited with permission.




                                                                                                     29

Table 1: Percentage of adult women reporting sexual victimization by an intimate
partner: population-based surveys
Country       Study population       Year    Sample    Assaulted               Percentage ever assaulted
                                             size      (attempted or
                                                       completed)
                                                       in last 12 months %)
                                                                               Attempted or         Completed
                                                                               completed            forced sex (%)
                                                                               forced sex (%)
Brazil        Sao Paulo   a          2000    941       2.8                     10.1
              Pernambuco    a        2000    1188      5.6                     14.3
Canada        National               1993    12,300                            8.0
              Toronto                1991-   420                               15.3  b

                                     1992
Chile         Santiago               1997    310       9.1
Finland       National               1997-   7,051     2.5                     5.9
                                     1998
Japan         Yokohama     a         2000    1,287     1.3                     6.2
Indonesia     Central Java           1999-   765       13.0
                                     2000
México        Durango                1996    384                               42.0
              Guadalajara            1996    650       15.0                    23.0
Nicaragua     León                   1993    360                               21.7
              Managua                1997    378       17.7
Peru          Lima  a                2000    1,086     7.1                     22.5
              Cusco  a               2000    1,534     22.9                    46.7
Puerto Rico   National               1993-   7,079                                                  5.7b

                                     1996
Sweden        Teg, Umea              1991    251                               7.5  c

Switzerland   National               1994-   1,500                             11.6
                                     1995
Thailand      Bangkok   a            2000    1,051     17.1                    29.9
              Nakornsawan    a       2000    1,027     15.6                    28.9
Turkey        East and southeast     1998    599                                                    51.9b

              Anatolia
United        England, Scotland      1989    1,007                                                  14.2d

Kingdom       and Wales
              North London           1993    430       6.0 b                   23.0  b

United        National               1995-   8,000     0.2 b                   7.7  b

States                               1996
West Bank     Palestinians           1995    2,410     27.0
and Gaza
Strip
Zimbabwe      Midlands Province      1996    966                               25.0
aPreliminary results from WHO multi-country study on women´s health and domestic violence. Geneva:
WHO. Sample size reported is denominator for the prevalence rate and not the total sample size of the
study.
bSample group included women who had never been in a relationship and therefore were not at risk of
being assaulted by an intimate partner.
coffenders reported to be husbands, boyfriends and acquaintances.
dweighted estimate; unweighted prevalence rate was 13.9%

Source: WHO, World Report on Violence and Health.




                                                                                                      30

Table 2: Percentage of women aged 16 years and older who report having been
sexually assaulted in previous 5 years: selected cities, 1992-1997

Country               Study population       Year         Sample      Percentage of
                                                          Size        women
                                                                      sexually
                                                                      assaulted in
                                                                      previous 5
                                                                      years (%)
Africa
Botswana              Gaborone               1997         644         0.8
Egypt                 Cairo                  1992         1,000       3.1
South Africa          Johannesburg           1996         1,006       2.3
Tunisia               Grand-Tunis            1993         1,087       1.9
Uganda                Kampala                1996         1,197       4.5
Zimbabwe              Harare                 1996         1,006       2.2

Latin America
Argentina             Buenos Aires           1996         1,000       5.8
Bolivia               La Paz                 1996         999         1.4
Brazil                Rio de Janeiro         1996         1,000       8.0
Colombia              Bogotá                 1997         1,000       5.0
Costa Rica            San José               1996         1,000       4.3
Paraguay              Asunción               1996         587         2.7

Asia
China                 Beijing                1994         2,000       1.6
India                 Bombay                 1996         1,200       1.9
Indonesia             Jakarta and Surabaya 1996           1,400       2.7
Mongolia              Ulaanbaatar,           1996         1,201       3.1
                      Zuunmod
Philippines           Manila                 1996         1,500       0.3

Eastern Europe
Albania               Tirana                 1996         1,200       6.0
Hungary               Budapest               1996         756         2.0
Lithuania             Diauliai, Kaunas,      1996         1,201       3.1
                      Klaipeda,
                      Panevezys, Vilnius

Source: International crime victimization survey (UNICRI), cited in WHO World Report
on Violence and Health.




                                                                                  31

Table 3: Comparison of advantages and disadvantages of GBV costing and impact
estimation methodologies

                                  Issues using methodology            Properties of results
Monetary costing:
Accounting (direct costs)         -  Social norms may inhibit the    - Allow dimensioning of
                                     use of medical, mental health,    GBV relative to other
                                     police, and justice services by   issues
                                     women victims of GBV            - Can be used to
                                  -  Information on usage of           determine cost
                                     medical, mental health, police,   effectiveness of
                                     and justice services by victims   preventing GBV
                                     is rarely available or accurate - Inform allocation of
                                  -  Information on the unit cost of   public funds
                                     medical, mental health, police, - Tend to underestimate
                                     and justice services is rarely    the cost of GBV
                                     available or subject to severe    (inhibited demand,
                                     distortion                        productivity losses,
                                  -  Willingness to pay for a life     and intergenerational
                                     free of GBV may not be            transmission of
                                     reflected by government's         violence are not
                                     allocation of public funds        included)
                                  -  Cost categories are essentially
                                     arbitrary
Estimation of indirect costs      -  Special large sample surveys    - Allow dimensioning of
(productivity and earnings losses    that include both GBV data        GBV relative to other
of victims and perpetrators)         and labor data are expensive      issues. However,
                                  -  GBV data gathering requires       indirect costs estimates
                                     extensive training of survey      for other epidemics
                                     personnel and safeguards for      may not be available
                                     the privacy and safety of the     for some countries
                                     respondent                      - Can be used to
                                  -  Special violence surveys,         determine cost
                                     violence modules in DHS, and      effectiveness of
                                     victimization surveys have        preventing GBV
                                     been recently conducted in      - Inform allocation of
                                     many countries. Some surveys      public funds
                                     contain data on labor force     - Hard to know if they
                                     participation and earnings        underestimate or
                                  -  Methodologies can be              overestimate the cost
                                     complex                           of GBV
                                  -  Econometric problems yet to
                                     be resolved (simultaneity)




                                                                                  32

Non-monetary costing:
Disability adjusted life years lost - Information on premature       - Allow dimensioning of
(DALYs)                               death, physical injuries, and    GBV relative to other
                                      mental secuelae caused by        issues. However, e
                                      GBV may be hard to obtain or     DALYs estimates for
                                      estimate                         other epidemics may
                                    - Public health data to estimate   not be available
                                      DALYs for conditions listed    - Can be used to
                                      above may not be available       determine cost
                                                                       effectiveness of
                                                                       preventing GBV
                                                                     - Inform allocation of
                                                                       public funds
                                                                     - Estimation of the
                                                                       health impact of GBV
                                                                       is very accurate, but
                                                                       does not include other
                                                                       costs (use of judicial
                                                                       system and
                                                                       productivity losses)
Impact on health and education      - Both large and small sample    - Allow dimensioning of
outcomes ­Naïve comparisons           surveys can be used              GBV relative to other
                                    - GBV data gathering requires      issues. However,
                                      extensive training of survey     impact estimates for
                                      personnel and safeguards for     other epidemics may
                                      the privacy and safety of the    not be available
                                      respondent                     - Hard to know if they
                                    - Special violence surveys,        underestimate or
                                      violence modules in DHS, and     overestimate the cost
                                      victimization surveys have       of GBV
                                      been recently conducted in
                                      many countries. Generally,
                                      these surveys contain data on
                                      health and education outcomes
                                      for all household members
                                    - Simple to calculate
                                    - Control group differs from
                                      victims group in more aspects
                                      than the experience of
                                      violence
                                    - Econometric problems
                                      (simultaneity)




                                                                                  33

Impact on health and education  - Special large sample surveys  - Allow dimensioning of
outcomes ­Propensity Score        that include GBV data are       GBV relative to other
Matching                          expensive                       issues. However,
                                - GBV data gathering requires     impact estimates for
                                  extensive training of survey    other epidemics may
                                  personnel and safeguards for    not be available
                                  the privacy and safety of the - Hard to know if they
                                  respondent                      underestimate or
                                - Special violence surveys,       overestimate the cost
                                  violence modules in DHS, and    of GBV
                                  victimization surveys have
                                  been recently conducted in
                                  many countries. Generally,
                                  these surveys contain data on
                                  health and education outcomes
                                  for all household members
                                - Complex estimation
                                  methodologies
                                - Econometric problems yet to
                                  be resolved (simultaneity)




                                                                            34

Table 4: Prevalence of gender based violence in Peru (2000), Haiti (2000), and
Zambia (2002) ­ Currently married or co-habitating women
                                    Peru             Haiti               Zambia
Gender Based Violence           N*        %      N         %          N         %
Type
Ever experienced              12,522 39.80     2,260 14.47          3,487 41.76
physical violence by
partner
Currently experiencing        n.a.    n.a.     2,210 12.53          1,456 23.03
physical violence by
partner**
Ever experienced sexual       n.a.    n.a.     2,260 18.89          3,485 5.5
violence by partner
Currently experiencing        n.a.    n.a.     2,232 17.43          452     6.6
sexual violence by
partner
Ever experienced              12,495 26.63     2,256 2.93           3,479 21.96
physical violence by
relatives (other than
husband in the
household)
Ever experienced              n.a.    n.a.     2,256 0.58           n.a.    n.a.
physical violence by
strangers
Ever experienced sexual       n.a.    n.a.     n.a.     n.a.        3,490 13.00
violence by other than
husband
Notes: * N= Number of women who answered questions regarding each type of gender based
violence
      **Current = in the last 12 months




                                                                                    35

   Table 5: Risk factors for intimate partner violence

     Individual-level43           Relationship-level44      Community-level45          Societal level46

     Socialization and learning
     witnessing intimate                                         Neighborhood          cultural norms that
     partner violence as a                                  crime rates (+)            support violence as an
     child (+)47                                                                       accepted way to
                                                                                       resolve conflicts or to
                                                                                       punish
                                                                                       transgressions(+)
     suffering abuse as a         Association with          Absent or maladaptive
     child (+)                    gang, delinquent or       teaching of alternatives
                                  patriarchal peers (+)     to violence (+)
     Power relations and patriarchal gender norms
     absent or rejecting father   male control of           Norms that support         norms that support
     (+)                          household decision-       male dominance over        male dominance over
                                  making and wealth         women and that             women and that
                                  (+)                       require women's            require women's
                                                            obedience and sexual       obedience and sexual
                                                            availability (+)           availability (+)

                                  Controlling behavior      Policies and laws that     Policies and laws that
                                  on the part of the        discriminate against       discriminate against
                                  husband (+)               women in social,           women in social,
                                                            economic and political     economic and political
                                                            spheres                    spheres
                                  Multiple
                                  partners/wives for the
                                  husband (+)
                                  Differences in
                                  spousal age and
                                  education (+)
                                  Household structure       Area of residence
                                  (nuclear versus non-      (region, urban, rural,
                                  nuclear) (+/ -) 48        large cities, small
                                                            cities)49




43Individual level: biological and personal history factors among both victims and perpetrators
44Relationship level: proximal social relationships, including relations with friends, peers and family.
45Community level: community context in which social relationships are embedded, including schools,
workplaces and neighborhoods
46Societal level: larger societal factors that "create an acceptable climate for violence, reduce inhibitions
against violence, create an sustain gaps between segments of society (WHO, 2002: 13)."
47For boys, witnessing violence increases the risk of becoming an abuser, whereas for girls it increases the
risk for future victimization.
48Kishor and Johnson (2004) find that women in nuclear households experience a higher probability of
violence.
49Kishor and Johnson (2004) find that women in rural areas have a lower probability of experiencing
violence compared to women living in rural areas.


                                                                                                             36

Human capital and employment
female educational level    economic hardship         Lack of economic
(-)                         (+)                       opportunities for men
                                                      (+)
Male educational level                                                            access and control over
(-)                                                                               economic resources for
                                                                                  women (+/-)
Women engaged in
income generation
activities (+/-)

Life cycle
age of woman (-)            length of relationship
                            (-)
Triggers
HIV status of man or        alcohol and substance
woman (+)                   abuse by male (+)
Source: Morrison, Ellsberg and Bott (2004)
Note: Signs in parenthesis indicate the sign of the correlation between the risk factor and the
probability of experiencing intimate partner violence.




                                                                                                      37

Table 6: Logit estimates (odds ratios) of risk factors of ever experiencing physical
violence by intimate partner
Explanatory Variable                           Peru         Haiti      Zambia

Individual Characteristics
Age group (ref: age group 15-19)
20-24                                               1.2154 -           1.15223
25-29                                            1.7405*** 0.5692      1.31837
30-34                                            1.9305*** 0.5237      1.05689
35-39                                            1.9000*** 0.5949      0.98928
40-44                                            2.1057*** 0.6074      1.12497
45-49                                            2.3565*** 0.2427**    0.85898
Woman's age at marriage (ref: less
than 15 years old)
15-19                                            0.8125*** -           0.80417**
20-24                                            0.5673*** -           0.76903*
25+                                              0.3778*** -           0.60192*
Age at first intercourse                                  - 0.9065**   -
Number of unions (ref: one)
More than one union                                0.8894* 0.8024      0.59074***
Number of children ever born (ref:
no children)
1-2                                              1.6734*** -           1.34279*
3-4                                              1.8973*** -           1.22302
5+                                               2.1348*** 1.0720      1.36924
Woman's education level (ref:
none)
Primary                                             1.0898 -           1.16239
Secondary or higher                                 1.0557 -           0.90155
Woman was hurt by father or                      1.5114*** 2.2599**    n.a.
punished physically as a child
Partner's Characteristics
Partner's education level (ref:none)
Primary                                             1.2062 0.8028      0.92185
Secondary or higher                                 1.3137 1.1896      0.98975
Husband's drunkenness (ref: does
not drink)
Never comes home drunk                                 Ref 1.1323      n.a.
Sometimes comes home drunk                       2.0344*** 2.8432*** n.a.
Frequently comes home drunk                      9.2491*** 3.9657*** n.a.

Relationship-level
Spousal age difference (ref:
husband is younger)
Husband is 0-4 years older                        0.8628** 0.7773      0.79712
Husband is 5-9 years older                        0.8602** 0.4559**    0.65368*
Husband is 10-14 years older                     0.6861*** 0.6960      0.68508
Husband is 15+ years older                       0.6634*** 0.5577      0.54586**




                                                                                 38

Spousal education difference (ref:
husband has less education)
Both have no education                                       - -             0.85540
Both have same level of education                    0.7222*** -             1.04217
Husband has more education                           0.7835*** 1.5768*       0.97305
Marital duration (ref: 0-4 years)                       1.1315
5-9 years                                                    - 1.7118        1.70546***
10-14 years                                                  - 1.2287        1.7422***
15 + years                                                 Ref 1.7263        2.0125***
Control by partner (ref: non-                                - 1.9035*** n.a.
controlling partner)
Decision making participation (ref:                    0.9206* -             0.8986
does not participate in economic
decisions)
Family structure (ref:nonnuclear)
Nuclear                                                 1.0582 1.6436        1.2471**
Household wealth status (ref:
poorest quintile)
Second quintile                                       1.1838** 1.3066        -
Middle quintile                                      1.2222*** 0.9467        -
Fourth quintile                                         1.0164 1.1531        -
Richest quintile                                     0.7104*** 0.8072        -

Community-level
Area of residence (ref: rural)
Urban                                                1.3042*** 1.9499**      1.3831***


- Log likelihood                                       7107.63 345.40        2277.59
Number of Women                                          11539 877           3450

LR chi squared (29)                                    1347*** 79.32***      134.26***

Pseudo R squared                                        0.0866 0.1030        0.0286
Percentage predicted correctly                         65.53% 84.61%         59.94%
Percentage predicted correctly                         59.67% 84.0%          58.20%
naïve model
Notes: ref= omitted category ; n.a.= not available in this dataset ; - = dropped due to
collinearity or an alternative proxy variable that performed better; * significant at 10%;
**significant at 5%; *** significant at 1%




                                                                                        39

Table 7: Logit estimates (odds ratios) of risk factors of currently experiencing
physical violence by intimate partner
Explanatory Variable                  Haiti     Zambia

Individual Characteristics
Age group (ref: age group 15-19)
20-24                                0.4151*    0.71582
25-29                                0.2094*** 0.52332*
30-34                                0.1896*** 0.57384
35-39                                0.1969**   0.42564*
40-44                                0.1809**   0.63226*
45-49                                0.0888*** 0.27863**
Woman's age at marriage (ref: less
than 15 years old)
15-19                                -          0.97139
20-24                                -          0.96351
25+                                  -          0.91454
Age at first intercourse             0.9311     -
Number of unions (ref: one)
More than one union                  0.7290     2.71248***
Number of children ever born
1-2                                  -          0.66643
3-4                                  -          0.62293
5+                                   0.8899     0.51987*
Woman's education level (ref:
none)
Primary                              -          0.77232
Secondary or higher                  -          0.77858
Woman was hurt by father or          2.3301*** n.a.
punished physically as a child
Partner's Characteristics
Partner's education level (ref:none)
Primary                              0.7545     0.62608
Secondary or higher                  1.4673     0.60409
Husband's drunkenness (ref: does
not drink)
Never comes home drunk               1.1800     n.a.
Sometimes comes home drunk           3.2994*** n.a.
Frequently comes home drunk          5.3399*** n.a.

Relationship-level characteristics
Spousal age difference (ref:
husband is younger)
Husband is 0-4 years older           0.7527     0.70673
Husband is 5-9 years older           0.4255**   0.75732
Husband is 10-14 years older         0.5938     0.61200
Husband is 15+ years older           0.5743     0.73365




                                                                                  40

Spousal education difference (ref:
husband has less education)
Both have no education                -            0.52037
Both have same level of education     -            1.15421
Husband has more education            1.4180       0.88489
Marital duration (ref: 0-4 years)
5-9 years                             2.3576**     0.57037***
10-14 years                           2.4616       0.34302***
15 + years                            2.8136*      0.31774***
Control by partner (ref : non-        1.8870*** n.a.
controlling partner)
Decision making participation (ref:   -            0.87148
does not participate in economic
decisions)
Family structure (ref:nonnuclear)
Nuclear                               1.9318*      0.95762
Household wealth status (ref:
poorest quintile)
Second quintile                       1.2325       -
Middle quintile                       0.7018       -
Fourth quintile                       1.0038       -
Richest quintile                      0.6340       -

Community-level
Area of residence (ref: rural)
Urban                                 2.3357*** 1.13419


- Log likelihood                      307.78       895.99
Number of Women                       859          1442

LR chi squared (29)                   86.45***     192.66***

Pseudo R squared                      0.1232       0.0971
Percentage predicted correctly        86.26%       65.33%
Percentage predicted naïve            85.80%       54.99%
model
Note: ref= omitted category ; n.a.= not available in this dataset ; - = dropped due to
collinearity or an alternative proxy variable that performed better; * significant at 10%;
**significant at 5%; *** significant at 1%




                                                                                        41

Table 8: Logit estimates (odds ratios) of risk factors of ever experiencing sexual
violence by intimate partner


Explanatory Variable                Haiti      Zambia
Individual Characteristics
Age group (ref: age group 15-19)
20-24                                -         1.1068
25-29                                0.1961*** 0.9382
30-34                                0.4534    1.3034
35-39                                0.4523    1.3357
40-44                                0.1941**  1.0060
45-49                                0.2507*   0.8478
Woman's age at marriage (ref: less
than 15 years old)
15-19                                -        -
20-24                                -        -
25+                                  -        -
First intercourse age                0.9379    0.8810***
Number of unions (ref: one)
More than one union                  0.5616**  0.8698
Number of children ever born
1-2                                  -         0.5678*
3-4                                  -         0.6350
5+                                   0.7409    0.3649**
Woman's education level (ref:
none)
Primary                              1.3262    0.5813*
Secondary or higher                  -         0.8981
Woman was hurt by father             1.4133   n.a.
Partner's Characteristics
Partner's education level (ref:none)
Primary                              0.6255    1.2925
Secondary or higher                  0.6236    0.9284
Husband's drunkenness (ref: does
not drink)
Never comes home drunk               1.4644    n.a.
Sometimes comes home drunk           2.6995*** n.a.
Frequently comes home drunk          0.7993    n.a.
Relationship-level
Spousal age difference (ref:
husband is younger)
Husband is 0-4 years older           0.5749    2.5719
Husband is 5-9 years older           0.4351**  2.4575
Husband is 10-14 years older         0.7494    2.7036
Husband is 15+ years older           0.4509*   1.7344




                                                                               42

Spousal education difference (ref:
husband has less education)
Both have no education                -            0.9929
Both have same level of education     -            0.8146
Husband has more education            1.0283       0.9646
Marital duration (ref: 0-4 years)
5-9 years                             0.4611       0.8957
10-14 years                           0.7104       0.8602
15 + years                            0.8795       1.1689
Control by partner (ref: non-         2.1580*** n.a.
controlling partner)
Decision making participation (ref:   -            0.9621
does not participate in economic
decisions)

Family structure (ref:nonnuclear)
Nuclear                               2.8441       1.1055
Household wealth status (ref:
poorest quintile)
Second quintile                       0.8133       -
Middle quintile                       2.2735**     -
Fourth quintile                       0.8389       -
Richest quintile                      0.7253       -
Community-level
Area of residence (ref: rural)
Urban                                 1.9024*      1.8285***

- Log likelihood                      226.157      488.44
Number of Women                       516          2299

LR chi squared (29)                   60.76***     39.34*

Pseudo R squared                      0.1184       0.0387
Percentage predicted correctly        80.62%       94.21%
Percentage predicted correctly        80.23%       94.215%
naïve model
Note: ref= omitted category ; n.a.= not available in this dataset ; - = dropped due to
collinearity or an alternative proxy variable that performed better; * significant at 10%;
**significant at 5%; *** significant at 1%




                                                                                        43

Table 9: Logit estimates (odds ratios) of risk factors of currently experiencing sexual
violence by intimate partner

Explanatory Variable                Haiti
Individual Characteristics
Age group (ref: age group 15-19)
20-24                                6.9351**
25-29                                1.1960
30-34                                2.9784**
35-39                                2.9181**
40-44                                1.0127
45-49                                -
Woman's age at marriage (ref: less
than 15 years old)
15-19                                -
20-24                                -
25+                                  -
First intercourse age                0.9558
Number of unions (ref: one)
More than one union                  0.6347
Number of children ever born
1-2                                  -
3-4                                  1.5953
5+                                   -
Woman's education level (ref:
none)
Primary                              1.2670
Secondary or higher                  -
Woman was hurt by father             0.9045
Partner's Characteristics
Partner's education level (ref:none)
Primary                              0.7071
Secondary or higher                  0.8136
Husband's drunkenness (ref: does
not drink)
Never comes home drunk               1.4531
Sometimes comes home drunk           3.0503***
Frequently comes home drunk          1.0443
Relationship-level
Spousal age difference (ref:
husband is younger)
Husband is 0-4 years older           0.5772
Husband is 5-9 years older           0.4507**
Husband is 10-14 years older         0.7300
Husband is 15+ years older           0.4288*




                                                                                    44

Spousal education difference (ref:
husband has less education)
Both have no education                -
Both have same level of education     -
Husband has more education            1.0150
Marital duration (ref: 0-4 years)
5-9 years                             0.5522
10-14 years                           0.8502
15 + years                            1.1429
Control by partner (ref: non-         2.2035
controlling partner)

Family structure (ref:nonnuclear)
Nuclear                               3.1156**
Household wealth status (ref:
poorest quintile)
Second quintile                       0.9217
Middle quintile                       2.1432**
Fourth quintile                       0.6973
Richest quintile                      0.5543

Community level
Area of residence (ref: rural)
Urban                                 2.2994**
- Log likelihood                      208.87
Number of Women                       509

LR chi squared (29)                   63.27***

Pseudo R squared                      0.1315
Percentage predicted correctly        82.32%
Percentage predicted correctly        81.93%
Naïve model
Note: ref= omitted category ; n.a.= not available in this dataset ; - = dropped due to
collinearity or an alternative proxy variable that performed better; * significant at 10%;
**significant at 5%; *** significant at 1%




                                                                                        45

Table 10: Logit estimates (odds ratios) of risk factors of ever experiencing physical
violence by relatives

Explanatory Variable                            Peru         Haiti      Zambia
Individual Characteristics
Age group (ref: age group 15-19)
20-24                                              0.7698** 0.2976*     0.74093*
25-29                                                0.8378 0.1839*     0.84268
30-34                                              0.7635** 0.1426*     0.91493
35-39                                              0.7187** 0.2437      0.65262
40-44                                              0.7444** -           0.56197
45-49                                            0.6593*** 0.2474       0.61558
Woman's age at marriage (ref: less
than 15 years old)
15-19                                                      - -          1.0268
20-24                                                      - -          1.0795
25+                                                        - -          1.4341
First intercourse age                                      - 0.9678     -
Number of unions (ref: one)
More than one union                              2.3239*** 0.9104       1.7389***
Number of children ever born
1-2                                                  1.0073 -           0.7998
3-4                                                  0.9742 -           0.6431**
5+                                                   1.0535 0.3482*     0.6184**
Woman's education level (ref:
none)
Primary                                              1.0389 -           1.1440
Secondary or higher                                  1.0864 3.1595      1.5516**
Woman was hurt by father                                   - 1.0579     n.a.
Partner's characteristics
Partner's education level (ref:none)
Primary                                              0.8810 1.3258      0.9469
Secondary or higher                                  0.8829 1.6684      0.9394
Husband's drunkenness (ref: does
not drink)
Never comes home drunk                                     - 2.8400*    n.a.
Sometimes comes home drunk                         1.1206** 1.0194      n.a.
Frequently comes home drunk                      1.7536*** -            n.a.
Relationship-level
Spousal age difference (ref:
husband is younger)
Husband is 0-4 years older                                 - 0.6939     0.7378
Husband is 5-9 years older                                 - 0.9461     0.6547
Husband is 10-14 years older                               - 0.4594     0.6493
Husband is 15+ years older                                 - 0.4596     0.8797
Spousal education difference (ref:
husband has less education)
Both have no education                                     - -          0.8807
Both have same level of education                          - -          0.9351
Husband has more education                                 - 1.0633     1.1013



                                                                                  46

Marital duration (ref: 0-4 years)
5-9 years                                                     - 2.7755      0.8880
10-14 years                                                   - 3.8295      0.8143
15 + years                                                    - 3.4557      0.8594
Control by partner (ref: non-                                 - 1.3055      n.a.
controlling partner)
Decision making participation (ref:                           - -           1.1153
does not participate in economic
decisions)
Family structure (ref:nonnuclear)
Nuclear                                                       - 0.4416       0.9237
Household wealth status (ref:
poorest quintile)
Second quintile                                      0.6797*** 0.6222       -
Middle quintile                                        0.8869* 2.2189       -
Fourth quintile                                      0.5208*** 1.9779       -
Richest quintile                                     0.7966*** 2.7587        -
Community -level
Area of residence (ref: rural)
Urban                                                  1.1604** 0.8482       1.2800**
- Log likelihood                                       6535.28 122.67       1761.97
Number of Women                                          11558 767          3442

LR chi squared (29)                                  368.76*** 33.03        105.21***

Pseudo R squared                                         0.0274 0.1187      0.029
Percentage predicted correctly                          73.15% 95.6%        77.89%
Percentage predicted correctly                          73.19% 95.6%        77.98%
naïve model
Note: ref= omitted category ; n.a.= not available in this dataset ; - = dropped due to
collinearity or an alternative proxy variable that performed better; * significant at 10%;
**significant at 5%; *** significant at 1%




                                                                                        47

Table 11 : Mean and standard deviation­ education, health , and employment
outcomes - currently married or cohabitating women

Outcome                  Peru             Haiti                 Zambia
Women's health
Weight for Height        12810.72         9462.33               8934.56
(centimeters x           (7036.39)        (1754.58)             (1407.93)
kilograms)
Anemia (severity         0.747            3.24                  n.a.
degree)                  (0.82)           (0.80)
Number of Children       3.31             3.88                  3.49
                         (2.40)           (2.91)                (2.40)
Terminated Pregnancies   0.207            0.174                 0.2187
                         (0.405)          (0.38)                (0.413)
Last Child Wanted        1.87             1.94                  1.58
(index 1 =wanted ­       (0.87)           (0.86)                ( 0.78)
3=did not want more
children)
Sexually Transmitted     20.65            3.4                   2.5
Disease (%)              (40.48)          (18.0)                (15.56)
Genital Sore /Ulcers (%) n.a.             4.36                  3.17
                                          (20.0)                (17.54)
Delivery Complications   36.41            n.a.                  n.a.
(% of women)             (48.12)
Women's use of health
services
Visited Health Facility  47.65            41.0                  73.85
(%)                      (50.0)           (49)                  (44)
Antenatal Care (%)       97.63            79.0                  93.66
                         (15.21)          (40)                  (24.4)
Births Assisted by       53.11            56.0                  40.88
Health Care Professional (49.91)          (50)                  (49.17)
(%)
Contraceptive Use (%)    87.56            42.5                  50.8
                         (33.01)          (49.0)                (50.0)
Unmet Family Planning    14.58            60.1                  47.13
Needs (%)                (35.30)          (49.0)                (49.93)
Women's employment
Employed and Earning     0.4624           0.5948                0.5712
Cash (probability)       0.4599           (0.4568)              (0.4950)
Children's health
(ages 0-5)
Diarrhea (%)             16.20            26                    21.9
                         (0.34)           (40.0)                (37.0)
Coughing (%)             n.a.             67                    39.3
                                          (44.0)                (45.0)


                                                                            48

Anemia (%)                 54.77          93              n.a.
                           (49.80)        (79.5)
Height for Age             2244.8         2760.61         1721.13
(centimeters x age in      (2506.53)      (2669.78)       (2529.94)
months)
Weight for Height          6132.54        4297.37         4524.99
(centimeters x             (2755.175)     (2716.66)       (2698.82)
kilograms)
Immunization (%)           36.68          61              8.48
                           (48.21)        (47.0)          (17.0)
Under 5 mortality (per     68.46          75.27           108.15
1000 births)1              (232.12)       (221.22)        ( 259.93)
Children's educational
achievement (ages 7
and older)
Education Years            0.59           2.28            2.48
                           (1.16)         (2.33)          (2.15)
Education Gap              0.865          2.44            1.64
                           (0.28)         (2.30)          (2.19)
School Attendance (%)      47.8           49              55
                           (50.50)        (43.0)          (40.0)
% Mothers using            42.87                  n.a.    n.a.
violence to discipline     (50.21)
children
Note: Standard deviations in parenthesis




                                                                      49

Table 12: Average effects of lifetime physical violence by intimate partner

Outcome                  Peru               Haiti                 Zambia
Women's health
Weight for Height        158.317            -186.427              29.874
(centimeters x
kilograms)
Anemia (severity         -0.048**           0.384 *               n.a.
degree)
Number of Children       0.030              -0.221                0.081
Terminated Pregnancies   0.101***           0.057                 0.009
Last Child Wanted        0.090 points***    0.109                 -0.034
(index 1 =wanted ­
3=did not want more
children)
Sexually Transmitted     4.0 percentage     1.4 percentage points 1.9 percentage
Disease (%)              points ***                               points***
Genital Sore /Ulcers (%) n.a.               4.3 percentage        2.8 percentage
                                            points*               points***
Delivery Complications   7.0 percentage     n.a.                  n.a.
(% of women)             points***
Women's use of health
services
Visited Health Facility  1.4 percentage     7.9 percentage points 3.5 percentage
(%)                      points                                   points**
Antenatal Care (%)       -0.1 percentage    -13.9 percentage**    -1.3 percentage
                         points             points                points
Births Assisted by       5.2 percentage     1 percentage point    -0.3 percentage
Health Care Professional points***                                points
(%)
Unmet Family Planning    -2.9 percentage    3.5 percentage points 0.4 percentage
Needs (%)                points***                                points*
Contraceptive Use (%)    3.6 percentage     13.6 percentage       1.6 percentage
                         points***          points                points
Women's employment
Employed and Earning     0.062***           0.022*                0.060**
Cash (probability)
Children's health (ages
0-5)
Diarrhea (%)             6.7 percentage     6.2 percentage points 1.6 percentage
                         points***                                points
Coughing (%)             n.a.               9.4 percentage        0.6 percentage
                                            points                points
Anemia (%)               -2.7 percentage    19.2 percentage*      n.a.
                         points             points
Height for Age           -111.225 *         - 150.061             36.384


                                                                                  50

(centimeters x age in
months)
Weight for Height           -124.695             -143.67                  -69.323
(centimeters x
kilograms)
Immunization (%)            12.9 percentage      -8.3 percentage          -0.5 percentage
                            points***            points                   points
Under 5 mortality (per      -6.852               13.79                    10.646
1000 births) 1



Children's educational
achievement (ages 7
and older)
Education Years             n.a                  -0.254 of a year         0.062 of a year
Education Gap               -0.043 of a          0.182 of a year          0.033 of a year
                            year*
School Attendance (%)       2.2 percentage       -1.9 percentage          -2.7 percentage
                            points***            points                   points
% Mothers using             7.0 percentage       n.a.                     n.a.
violence to discipline      points***
children
* significance at 10% ** significance at 5% *** significance at 1%

1This is not a population child mortality rate. This variable is the sample average the
number of children under 5 who died divided by each woman's total number of births
per 1000. In this case both rates are similar since there is no significant difference in the
total number of children between the victims group and the control group.




                                                                                            51

Table 13: Average effects of current physical violence by intimate partner

Outcome                              Haiti
Women's health
Weight for Height (centimeters x     -205.576
kilograms)
Anemia (severity degree)            0.157
Number of Children                   0.180
Terminated Pregnancies              0.107 **
Last Child Wanted                    - 0.001
(index 1 =wanted ­3=did not want
more children)
Sexually Transmitted Disease (%)     2.4 percentage points
Genital Sore /Ulcers (%)             5.8 percentage points**
Women's use of health services
Visited Health Facility (%)          17.6 percentage points**
Antenatal Care (%)                   - 4.8 percentage points
Births Assisted by Health Care       11.4 percentage points
Professional (%)
Unmet Family Planning Needs (%)     8.0 percentage points
Contraceptive Use (%)                4.1 percentage points
Women's employment
Employed and Earning Cash            0.013
(probability)
Children's health (ages 0-5)
Diarrhea (%)                        -6.3 percentage points
Coughing (%)                        -3.2 percentage points
Anemia (%)                          2.24 percentage points*
Height for Age (centimeters x age    322.29
in months)
Weight for Height (centimeters x     -114.734
kilograms)
Immunization (%)                    -0.7 percentage points
Under 5 mortality (per 1000 births)1 39.279
Children's educational
achievement (ages 7 and older)
Education Years                     -0.518 of a year*
Education Gap                       0.120 of a year
School Attendance (%)                3.2 percentage points
% Mothers using violence to          n.a.
discipline children
* significance at 10% ** significance at 5% *** significance at 1%




                                                                              52

Table 14: Average effects of lifetime sexual violence by intimate partner

Outcome                                  Haiti
Women's health
Weight for Height (centimeters x         351.941 **
kilograms)
Anemia (severity degree)                 -0.025
Number of Children                       0.251
Terminated Pregnancies                   0.097**
Last Child Wanted (index 1 =wanted ­     0.133
3=did not want more children)
Sexually Transmitted Disease (%)         2.2 percentage points
Genital Sore /Ulcers (%)                 4.9 percentage points*
Women's use of health services
Visited Health Facility (%)              13.9 percentage points***
Antenatal Care (%)                       3.8 percentage points
Births Assisted by Health Care           8.6 percentage points
Professional (%)
Unmet Family Planning Needs (%)          -5.2 percentage points
Contraceptive Use (%)                    18.8 percentage points
Women's employment
Employed and Earning Cash                0.067**
(probability)
Children's health (ages 0-5)
Diarrhea (%)                             - 2.4 percentage points
Coughing (%)                             - 7.0 percentage points
Anemia (%)                               6.5 percentage points
Height for Age (centimeters x age in     188.943 points
months)
Weight for Height (centimeters x         335.127
kilograms)
Immunization (%)                         - 8.3 percentage points
Under 5 mortality (per 1000 births)1     27.864
Children's educational achievement
(ages 7 and older)
Education Years                          0.157 of a year
Education Gap                            -0.109 of a year
School Attendance (%)                    1.7 percentage points
% Mothers using violence to              n.a.
discipline children
* significance at 10% ** significance at 5% *** significance at 1%




                                                                             53

Table 15: Average effects of currently experiencing sexual violence by intimate
           partner

Outcome                             Haiti
Women's health
Weight for Height (centimeters x    308.229
kilograms)
Anemia (severity degree)            0.267
Number of Children                  -0.351
Terminated Pregnancies              -0.01
Last Child Wanted (index 1          0.312**
=wanted ­3=did not want more
children)
Sexually Transmitted Disease (%)    6.3 percentage points***
Genital Sore /Ulcers (%)            5.2 percentage points*
Women's use of health services
Visited Health Facility (%)         7.2 percentage points
Antenatal Care (%)                  -9.5 percentage points
Births Assisted by Health Care      -6.6 percentage points
Professional (%)
Unmet Family Planning Needs (%)     -3.6 percentage points
Contraceptive Use (%)               - 2.1 percentage points
Women's Employment
Employed and Earning Cash           0.000
(probability)
Children's health (aged 0-5)
Diarrhea (%)                        1 percentage point
Coughing (%)                        - 4.7 percentage points
Anemia (%)                          17.3 percentage points
Height for Age (centimeters x age   -560.38
in months)
Weight for Height (centimeters x    -426.927
kilograms)
Immunization (%)                    6.1 percentage points
Under 5 mortality (per 1000 births) - 4.843
1

Children's educational
achievement (ages 7 and older)
Education Years                     0.122 of a year
Education Gap                       0.020 of a year
School Attendance (%)               1.9 percentage points
% Mothers using violence to         n.a.
discipline children
* significance at 10% ** significance at 5% *** significance at 1%




                                                                                54

Table 16: Average effects of lifetime physical violence by relatives

Outcome                                                         Peru
Women's health
Weight for Height (centimeters x kilograms)                     447.497**
Anemia (severity degree)                                        -0.018
Number of Children                                              0.062**
Terminated Pregnancies                                          0.021**
Last Child Wanted                                               0.138 points***
(index 1 =wanted ­3=did not want more children)
Sexually Transmitted Disease (%)                                1.4 percentage points
Genital Sore /Ulcers (%)                                        n.a.
Delivery Complications (% of women)                             9.5 percentage points***
Women's use of health services
Visited Health Facility (%)                                     1.5 percentage points
Antenatal Care (%)                                              -0.3 percentage points
Births Assisted by Health Care Professional (%)                 -3.4 percentage points**
Unmet Family Planning Needs (%)                                 -0.2 percentage points
Contraceptive Use (%)                                           2.5 percentage points***
Women's employment
Employed and Earning Cash (probability)                         0.042***
Children's health ( 0-5)
Diarrhea (%)                                                    2.8 percentage points**
Coughing (%)                                                    n.a.
Anemia (%)                                                      - 1.5 percentage points
Height for Age (centimeters x age in months)                    -136.61 *
Weight for Height (centimeters x kilograms)         1           -179.517 *
Immunization (%)                                                -2.4 percentage points
Under 5 mortality (per 1000 births)      2                      7.913
Children's educational achievement (ages 7 and
older)
Education Years                                                 n.a.
Education Gap                                                   -0.038 of a year*
School Attendance (%)                                           2.4 percentage points**
% Mothers using violence to discipline children                 9.8 percentage points***
* significance at 10% ** significance at 5% *** significance at 1%
1 In the case of Peru , this variable measures the weight for height of the youngest child
2 This is not a population child mortality rate. This variable is the sample average the number of
children under 5 who died divided by each woman's total number of births per 1000. In this case
both rates are similar since there is no significant difference in the total number of children
between the victims group and the control group.




                                                                                                 55

References

Australian Institute of Criminology. 2002. "Australian Crime: Facts and Figures 2001."
Canberra: Australian Institute of Criminology.

Bloch, F. and Rao, V. 2002. "Terror as a Bargaining Instrument: A Case Study of Dowry
Violence in Rural India". American Economic Review 92(4): 1029-43

Blumel, D.K. et al., 1993. "Who Pays? The Economic Costs of Violence against
Women." Queensland: Women's Policy Unit, Office of the Cabinet.

Buvinic, M. and A. Morrison. 1999. Technical Notes on Violence Prevention.
Washington, D.C. : Inter-American Development Bank, Sustainable Development
Department.

Campbell, J. 2002a. "Health Consequences of Intimate Partner Violence." Lancet 359
(9314): 1331-6.

Campbell, J. et al. 2002b. "Intimate Partner Violence and Physical Health
Consequences." Archives of Internal Medicine 162: 1157-1163.

Coker, A.L. et al. 2002. "Physical and Mental Health Effects of Intimate Partner
Violence for Men and Women." American Journal of Preventive Medicine 23(4): 260-8.

Coker, A.L. et al. 2000a. "Intimate Partner Violence and Cervical Neoplasia." Journal of
Women's Health and Gender-Based Medicine 9(9): 1015-23.

Coker, A.L. et al. 2000b. "Physical Health Consequences of Physical and Psychological
Intimate Partner Violence." Archives of Family Medicine 9: 451-7.

CDC. 2003. "Costs of Intimate Partner Violence against Women in the United States."
Atlanta, GA: Centers for Disease Control and Prevention, National Center for Injury
Prevention and Control.

Dahlberg, L. 1998. Youth Violence in the United States: Majore Trends, Risk Factors
and Prevention Approaches." American Journal of Preventive Medicine 14(4): 259-72.

Dehejia, R. and S.Whaba. 1998. " Causal Effects in Nonexperimental Studies :
Reevaluating the Evaluation of Training Programs."Journal of the American Statistics
Association. 94(448). Applications and Case Studies.

Dube, S. et al. 2001. "Childhood Abuse, Household Dysfunction, and the Risk of
Attempted Suicide throughout the Life Span: Findings from the Adverse Childhood
Experiences Study." Journal of the American Medical Association 286(24): 3089-96.

Duvvury, N. et al. 2000. Domestic Violence in India: A Summary Report of Four
Records Studies. Washington, DC: International Center for Research on Women.


                                                                                       56

Ellsberg, M. et al. 2001. " Researching Domestic Violence Against Women:
Methodological and Ethical Considerations." Studies in Family Planning 32(1): 1-16.

Ellsberg, M. et al. 2000. "Candies in Hell: Women's Experiences of Violence in
Nicaragua." Social Science and Medicine 51(11): 1595-1610.

Ellsberg, M. et. al. 1999. "Domestic Violence and Emotional Distress Among Nicaraguan
Women: Results from a Population-Based Study." American Psychologist, 54: 30-36.

Farmer, A. and J. Tiefenthaler. 1999. "The Employment Effects of Domestic Violence."
Fayetteville, AK: University of Arkansas, Department of Economics: Working paper.

Godenzi, A. and C. Yodanis, 1998. "Report on the Economic Costs of Violence against
Women." Fribourg: University of Fribourg.

Greaves, L. et al. 1995. "Selected Estimates of the Costs of Violence against Women."
London, Ontario : Centre for Research on Violence against Women and Children.

Haddix A.C et al., 1996. Prevention Effectiveness : A Guide to Decision Analysis and
Economic Evaluation. New York: Oxford University Press.

Health Canada. 2002. "Violence against Women. Impact of Violence on Women´s
Health." Ottawa: Health Canada.

Heckman J. et al. 1999. The Economics and Econometrics of Active Labor Market
Programs. Handbook of Labor Economics. Ashenfelter, A. and D. Card, editors. Vol 3.

Heise, L., Ellsberg, M. and Gottemoeller, M. 1999. "Ending Violence Against Women."
Population Reports. Series L, No. 11. Baltimore: Johns Hopkins University School of
Public Health.

Heise, L. 1997. "Violence Against Women: An Integrated, Ecological Model." World
Bank Working Paper . Washington, DC: The World Bank.

Heise, L. et al. 1994. "Violence against Women : The Hidden Health Burden." World
Bank Discussion Paper #255. Washington, D.C.: The World Bank.

Hernández Rosete, D. 1998. "Pobreza urbana y violencia doméstica en hogares de la
Ciudad de México." Acta Sociológica (enero-abril): 25-43.

International Planned Parenthood Federation. 2000. "What is Gender-Based Violence?"
Basta! Spring 2000.

Jalan and Ravallion. 2003. "Estimating the Benefit Incidence of an Antipoverty Program
by Propensity ­Score Matching." Journal of Business and Economic Statistics. 21(1).
American Statistical Association.



                                                                                     57

Jewkes, R. 2002. "Risk factors for domestic violence: findings from a South African
cross-sectional study." Social Science and Medicine 55: 1603-1617.


Jewkes, R. 2000. "Violence against Women: An Emerging Health Problem."
International Clinical Psychopharmacology 15 (Suppl. 3) : S37-S45).

Kerr, R. and J. McLean, 1996. "Paying for Violence: Some of the Costs of Violence
against Women in B.C." British Columbia: Ministry of Women´s Equality.

Kishor, S. and K. Johnson, 2004. " Profiling Domestic Violence ­ A Multi Country
Study." Measure DHS + Project. Maryland: ORC Macro.

Koenig, M. et al. 2003. "Women's Status and Domestic Violence in Rural Bangladesh:
Individual- and Community-Level Effects." Demography 40(2): 269-88.

Korf D.J. et al. 1997. "Economic Costs of Domestic Violence against Women." Utrecht,
Netherlands: Dutch Foundation of Women´s Shelters.

Larraín, S. 1999. Curbing Domestic Violence: Two Decades of Action. In: Morrison,
Andrew and María Loreto Biehl, Eds. Too Close to Home: Domestic Violence in the
Americas. Washington, D.C.: Inter-American Bank, 105-130.

Larraín, S. et al. 1997. Relaciones Familiares y Maltratro Infantil. Santiago, Chile:
UNICEF.

Laurence, L. and R. Spalter-Roth. 1995. "Measuring the Costs of Domestic Violence
against Women and the Cost-Effectiveness of Interventions: An Initial Assessment and
Proposals for Further Research." Washington, D.C.: Institute for Women's Policy
Research.

Leach, F. et al. 2003. "The Abuse of Girls in African Schools." London: DFID
Education Research Report No. 54.

Lloyd, S. 1999. "The Effects of Male Violence on Female Employment." Violence
against Women 5(4): 370-92.

Lozano, R. 1999. "The Impacts of Domestic Violence on Health: Mexico City," in
Morrison and Biehl (eds.), Too Close to Home: Domestic Violence in the Americas.
Washington, D.C.: Inter-American Development Bank.

Mansingh A. and P. Ramphal. 1993. "The Nature of Interpersonal Violence in Jamaica."
West Indian Medical Journal 42: 53-6.

Miller, T.R., M.A. Cohen and B. Wiersema, 1996. "Victim Costs and Consequences: A
New Look." National Institute of Justice Report. Washington, D.C.: National Institute
of Justice.


                                                                                       58

Molnar, B. et al. 2001. "Child Sexual Abuse and Subsequent Pathology: Results from the
National Comorbidity Survey." American Journal of Public Health 91(5): 753-60.

Morrison, A. and M.B. Orlando. 1999. "The Socioeconomic Costs of Domestic Violence:
Chile and Nicaragua, " in ," in Morrison and Biehl (eds.), Too Close to Home:
Domestic Violence in the Americas. Washington, D.C.: Inter-American Development
Bank.

Morrison, A., M. Ellsberg and S. Bott. 2004. "Addressing Gender-Based Violence in
Latin America and the Caribbean:A Critical Review of Interventions." Working Paper.
Washington, D.C.: World Bank.

Office of Women´s Policy. 1996. "The Financial and Economic Costs of Domestic
Violence in the Northern Territory." Northern Territory: KPMG.

Rao, V. 1998. "Wife-Abuse, Its Causes and Its Impact on Intra-Household Resource
Allocation in Rural Karnataka: A "Participatory" Econometric Analysis," in Gender,
Population, and Development, M.Krishnaraj, R.Sudarshan, A.Sharif (ed.), Oxford
University Press, Delhi and Oxford
Sánchez, F. et al. 2004. "Segunda Parte: Estudio de Mujeres", in Los costos de la
violencia intrafamiliar en Colombia. Bogotá: Universidad de los Andes, Centro de
Estudios sobre Desarrollo Económico.

Smith, J. 2000. "A Critical Survey of Empirical Methods for Evaluating Active Labor
Market Policies." Zeitschrift fur Volkswritschaft und Statistik 136(3): 1-22.

J. Smith and P. Todd. 2000. "Does Matching Overcome Lalonde's Critique of
Nonexperimental Estimators?" Working paper. London, Ontario: University of Western
Ontario.

Snively, S. 1994. "The New Zealand Economic Cost of Family Violence." Wellington,
New Zealand: Coopers and Lybrand.

Sorenson, S. 2003. "Funding Public Health: The Public´s Willingness to Pay for
Domestic Violence Prevention Programming." American Journal of Public Health 93:
1934-38.

Stanko, E.A. et al., 1998. "Counting the Costs : Estimating the Impact of Domestic
Violence in the London Borough of Hackney." Middlesex, UK: Brunel University.

Strauss, M. et al. 1980. Behind Closed Doors. New York: Doubleday.

Teutcsch, S.M. 1992. "A Framework for Assissing the Effectivenessof Disease and
Injury Prevention," Morbidity and Mortality Weekly Report 41 (RR-3): 1-13.




                                                                                    59

Thornberry, T.P. et al. 2001. "The Importance of Timing: The Varying Impact of
Childhood and Adolescent Maltreatment on Multiple Problem Outcomes." Development
and Psychopathology 13(4): 957-79.

United Nations. 1993. " Declaration on the Elimination of Violence Against Women."
United Nations General Assembly Resolution 48/104. NewYork: United Nations.

Windom, C.S. 1989. "The Cycle of Violence." Science 244: 160-166.

World Health Organization. 2004. "The Economic Dimensions of Interpersonal
Violence." Geneva: WHO, Department of Injuries and Violence Prevention.

Yodanis, C. et al. 2000. "The Benefits of Studying Costs: A Review and Agenda for
Studies on the Economic Costs of Violence against Women." Policy Studies 21, no. 3:
263-76.




                                                                                   60