A PROJECT REPORT

BANARAS HINDU UNIVERSITY

“Socioeconomic Determinants of Domestic

Violence in Karnataka, India: Evidence from the
National Family Health Survey (NFHS-5)”

SUBMITTED BY
SHALINI SINGH
MSc STATISTICS 2023-24
ROLL NO:23420STA050

SUPERVISOR
Prof. Manoj Kumar Chaudhary
DECLARATION
I hereby declare that the work presented in this research project titled ―Socioeconomic
Determinants of Domestic Violence: A Statistical Analysis‖ submitted in partial fulfilment of
M.Sc. (Statistics) at the Institute of Science, Banaras Hindu University, is an authentic record of
my original work carried out under the guidance of Dr. MKC, Institute of Science, BHU.

I confirm that this project has not been submitted for the award of any other degree.

Shalini Singh

M.Sc. Statistics

Institute of Science

Banaras Hindu University

Enrolment No: 450820

Roll No: 23420STA050

Date: 30 April, 2025

ACKNOWLEDGEMENT
I have put in sincere effort in this project. However, it would not have been possible without the
kind support and help of many individuals and I am deeply grateful to the Institute of Science,
BHU for providing me the opportunity to work on this project.

I am highly indebted to my supervisor, Prof. Manoj Kumar Chaudhary, for his guidance, support,
and constructive feedback throughout the project. His insights and encouragement were
invaluable.

Lastly, I am thankful to all unseen hands and minds who helped me directly or indirectly in
completing this study.

Shalini Singh
CONTENTS

SL. No. Topic Page No.

1. Objective & Introduction 5-6

2.
Review Of Literature 7-8

3.
Data & Methodology 9-10

4. Results 11-17

5. Discussion 18-19

6. Conclusion 20

7. References 21

8. Appendix 22

Objective

1. To determine the prevalence of domestic violence among married women of reproductive age
18-49 years.

2. To determine the types of domestic violence among these women.

3. To assess factors influencing domestic violence.

Abstract
Domestic violence remains a pervasive global and national public health concern with profound
implications for the well-being of individuals and societies. This dissertation investigates the
socioeconomic determinants of domestic violence, focusing on the context of India, where the
prevalence of such violence is alarmingly high. Utilizing statistical analysis of a comprehensive
dataset, this study examines the relationship between various socioeconomic factors, including
education, wealth, occupation, religion, residence, and others, and the likelihood of experiencing
domestic violence. The findings from multiple logistic regression models provide valuable
insights into the complex interplay of these factors, contributing to a deeper understanding of
vulnerability and risk. These results are contextualized within existing scholarly literature,
highlighting consistencies and discrepancies, and informing potential avenues for future research
and targeted interventions.

Introduction

Spousal violence, also known as domestic violence or intimate partner violence (IPV),
encompasses a cycle of harmful behaviour wherein an individual within a marital or cohabitating
relationship is subjected to emotional, physical, and/or sexual abuse. Different dimensions can be
utilized to categorize spousal violence, including the initiator and recipient of the violence, the
extent to which the violence is provoked by the victim, the nature of the harm inflicted, and so
forth. Yet, we believe that any discourse on family conflicts should encompass two distinct
aspects of violence: the societal endorsement of violence in a given scenario and the utilization
of violence for instrumental purposes. Although these two dimensions are continuous, for the
sake of clarity, we will simplify them into distinct categories.

Irrespective of race, gender, age, sexual orientation, or economic status, anyone can fall victim to
spousal violence. Broadly speaking, spousal violence can emerge when one partner exercises
control or coercion over the other using various methods. It's possible for one partner to be the
sole perpetrator, but in some cases, both partners may engage in abusive behaviour toward each
other through different means. The repercussions of spousal abuse extend beyond the immediate
victims and affect their families, friends, and communities. Indeed, the impact of spousal abuse
extends beyond the confines of any single home.

Abuse against women can be categorized into two main groups (Gordon, 2000). The first
category stems from programs that support survivors, addressing issues like sexual harassment
and domestic abuse, while the second is rooted in psychological and behavioural studies related
to sexual assault and family violence (Winstok, 2007). Spousal violence or intimate partner
violence is defined as "behaviour in a romantic relationship that inflicts or has the potential to
inflict physical, sexual, or psychological harm, such as acts of physical violence, sexual
harassment, psychological manipulation, and controlling behaviours" (Garcia-Moreno et al.,
2006, p. 1686).
Physical violence involves using physical contact to cause harm or suffering to an individual.
This includes actions like beating, punching, hitting, pulling, shoving, cutting, scraping, choking,
burning, and threats involving weapons like guns or knives. Sexual violence encompasses "any
sexual act, attempt to obtain a sexual act, unwelcome sexual remarks or advances, acts of
trafficking, or other coercive behaviour directed toward a person's sexuality by any individual,
regardless of their relationship to the victim, in any setting, including but not limited to homes
and workplaces" (Garcia-Moreno et al., 2006). In the context of intimate partner violence (IPV),
sexual abuse refers to pressuring a partner into engaging in sexual activities they find degrading,
injuring them during sex, or coercing them into sex without their consent (WHO, 2013).
Psychological abuse involves actions like threats, humiliation, withholding attention, and various
forms of control (e.g., isolating the victim, financial manipulation), characterized by demeaning
or degrading behaviour, often verbally (Maiuro & Eberle, 2008).

Spousal violence involves the abuse of power by a spouse, partner, ex-spouse, or ex-partner of
any gender, leading to a loss of autonomy, authority, and safety. This creates feelings of
helplessness and entrapment for victims, primarily women, who are subjected to repetitive acts
of physical, psychological, economic, emotional, verbal, and/or spiritual abuse. It also includes
repeated instances of coercing or pressuring women to witness violence perpetrated by their
partners against their children, other family members, friends, pets, or valued possessions
(Schwartz & Dekeseredy, 1997). Although it differentiates between survivors and aggressors and
recognizes the abusive misuse of power, it doesn't delve into the specific types of violence,
making assessment challenging (Winstok, 2007).

Domestic violence stands as a critical public health concern with far-reaching physical,
emotional, and social implications for victims. It affects individuals of all genders, ages, and
backgrounds, although women are disproportionately affected. Its impact ripples beyond
immediate victims, influencing children who witness such abuse and perpetuating cycles of
violence across generations. Tackling domestic violence necessitates a comprehensive approach
that involves not only supporting victims but also raising awareness, providing education, and
implementing legal measures to hold perpetrators accountable. Essential resources for those
enduring domestic violence include hotlines, shelters, counselling, and legal aid. Collaborative
efforts between law enforcement, social services, healthcare professionals, advocacy groups, and
community organizations are often required to prevent and address domestic violence, aiming to
create a safer environment for all individuals and break the cycle of violence that can persist
within families and relationships.

Rationale for the Study

Violence against women, particularly intimate partner violence (IPV), constitutes a significant
violation of human rights and a major impediment to achieving gender equality and overall
societal well-being. Globally, it is estimated that approximately one in three women has
experienced physical and/or sexual violence by an intimate partner. This widespread
phenomenon has been recognized by the World Health Organization (WHO) as a "global hidden
epidemic," underscoring its pervasive yet often concealed nature. The consequences of domestic
violence extend beyond immediate physical harm, encompassing long-term psychological
distress, reduced quality of life, and significant social and economic costs for individuals,
families, and communities.

In India, the issue of domestic violence is particularly acute. National surveys, such as the
National Family Health Survey (NFHS), consistently reveal a high prevalence of violence
against women. Approximately 30% of married women between the ages of 18 and 49 in India
report experiencing domestic violence. The most recent NFHS-5 (2019-21) data indicates that
around 32% of ever-married women in the country have experienced physical, sexual, or
emotional violence. Within India, the state of Karnataka has emerged as a region of particular
concern regarding spousal violence. Data from NFHS-5 shows a substantial increase in such
violence in Karnataka, with urban areas reporting a prevalence of 44.5%. Furthermore, a study
analyzing NFHS-5 data identified Karnataka as having the highest overall prevalence of
domestic violence in India, at 47.3%. Other reports corroborate these findings, indicating that
44% of married women in Karnataka experienced domestic violence between 2019 and 2021.

Understanding the factors that contribute to this high prevalence is crucial for developing
effective strategies to combat domestic violence. Research suggests that lower levels of
education and socioeconomic status are significant predictors of domestic violence. By
employing rigorous statistical analysis to explore the relationships between various
socioeconomic indicators and the occurrence of domestic violence, this study aims to provide
valuable insights into the underlying determinants of this critical social problem. The use of
nationally representative datasets like the NFHS, which employ robust sampling techniques ,
allows for a comprehensive examination of these associations. This dissertation will contribute to
the existing body of knowledge by analyzing a specific dataset to further elucidate the complex
interplay between socioeconomic factors and the risk of domestic violence.

Literature Review

Domestic violence (DV) is a deeply entrenched issue in both India and other developing
countries, drawing considerable attention from scholars, policymakers, and social activists. A
substantial body of literature has emerged seeking to understand the various facets of domestic
violence, particularly its relationship with socioeconomic and cultural factors. In the Indian
context, these investigations are critical for informing public policy and intervention strategies.

Theoretical Frameworks and Socioeconomic Inequality

The feminist theoretical framework provides a foundation for much of the discourse surrounding
domestic violence, emphasizing patriarchal structures that reinforce gender inequality and male
dominance. These frameworks propose that socioeconomic disparities exacerbate existing power
imbalances, increasing women's vulnerability to abuse within households and communities.
Gender-based socialization processes also contribute significantly, shaping norms that justify or
perpetuate violence against women.

Education and Domestic Violence

Education has consistently emerged as a crucial determinant of domestic violence. Lower

educational attainment in both women and their husbands is strongly associated with higher risk.
Women with no formal education are particularly vulnerable to emotional, physical, and sexual
abuse. However, the relationship is nuanced. Some studies suggest that higher education in
women may challenge traditional patriarchal norms, potentially triggering conflict and,
paradoxically, increasing the risk of violence in certain contexts. The educational levels of both
partners significantly influence marital dynamics and perceptions of gender roles.

Economic Status and Wealth Index

Socioeconomic status, often measured using the Wealth Index, is another major factor linked to
domestic violence. Women from poorer households exhibit higher odds of experiencing abuse.
Economic dependence tends to increase vulnerability, whereas financial independence can
reduce risk, albeit with complexity. In some cases, women who earn more than their husbands
report a higher incidence of violence, potentially due to a perceived threat to male authority and
traditional household hierarchies.

Employment and Occupation

The employment status of both women and their husbands also plays a significant role. Working
women may experience higher rates of domestic violence, possibly due to tensions arising from
role reversals or financial independence. Conversely, women whose husbands are unemployed
often face an increased risk of violence, likely due to economic stress and a perceived loss of
male identity. Research suggests that when women surpass their husbands economically, the
imbalance may lead to conflict and violence as a form of control or reassertion of dominance.

Religion and Regional Norms

Religion has been identified as a factor influencing domestic violence. Some studies suggest that
Muslim women may experience higher rates of violence, possibly due to specific cultural and
community norms. However, this finding requires nuanced interpretation to avoid stigmatization.
Similarly, the urban-rural divide reveals notable disparities, with women in rural areas often
facing a higher prevalence of abuse due to stronger patriarchal norms, limited support systems,
and greater social acceptance of violence.
Alcohol Consumption and Lifestyle Factors

The lifestyle of the husband, particularly alcohol consumption, has been consistently linked to
increased rates of domestic violence. Alcohol use can intensify conflicts, lower inhibitions, and
contribute to physical, emotional, and sexual abuse. Other lifestyle habits like smoking and drug
use have also been implicated.

Demographic and Marital Factors

Demographic elements such as age at marriage, parity (number of children), and family structure
significantly impact the occurrence of domestic violence. Early marriage exposes young women
to increased vulnerability due to their lack of autonomy and resources. A higher number of
children can elevate stress and economic strain, leading to a higher risk of violence.
Additionally, women who are divorced, widowed, or separated may experience continued or
elevated levels of abuse, influenced by social stigma and economic insecurity.

Empirical Findings and National Data (NFHS)

Data from national surveys like the National Family Health Survey (NFHS) provide vital
insights into the prevalence and patterns of domestic violence. The NFHS-5 findings reveal both
progress and persistent challenges. While a general declining trend in spousal violence is
observed in many states, regions like Karnataka, Telangana, Manipur, and Bihar still record
concerning levels—approximately 35%. Notably, Karnataka, along with Sikkim and Assam, has
witnessed an increase in spousal violence compared to NFHS-4, contrasting with reductions in
states like Andhra Pradesh and Meghalaya.

Focus on Karnataka

Karnataka serves as a microcosm of the broader national challenges. Despite being a hub of
technological advancement, particularly in Bangalore, the state continues to grapple with gender
inequality, regional disparities, and an urban-rural divide. These contradictions reflect in its
domestic violence statistics. Social hierarchies and ineffective governance often marginalize
women, limiting their access to justice and support.

Devaki Jain, drawing from her experience as a scholar-activist in Karnataka, critiques the
patriarchal foundations of public policy. She argues for a paradigm shift—from a "trickle-down"
to a "bubbling-up" approach—emphasizing the need for women‘s leadership in policymaking.
Rather than being passive recipients of aid, women must be empowered as active economic and
political agents. A broader strategic positioning is necessary, one that integrates women's issues
into the mainstream socio-economic discourse.
Data and Variables

Data Source

The data used for this analysis comprises a dataset derived from a survey focused on women's
health and socioeconomic characteristics. While the specific origin and collection methodology
of the dataset are not explicitly detailed in the provided information, it contains a range of
variables relevant to the study of domestic violence and its socioeconomic determinants. The
dataset includes responses from 9255 women, providing a substantial sample size for statistical
analysis.

Description of Variables

The analysis includes the following variables, with their original labels (where available) and the
renamed labels used for analysis, along with their descriptions and value labels:
 alcoholic_partner (SPSS Code: D113): This variable indicates whether the woman's partner
is alcoholic.

 0: No

 1: Yes

 age_group (SPSS Code: V012): This variable represents the age group of the woman.

 0: 16–29

 1: 30–49

 residence (SPSS Code: V025): This variable indicates the type of residence of the woman.

 0: Urban

 1: Rural

 highest_education (SPSS Code: V106): This variable represents the highest level of
education attained by the woman.

 0: No education

 1: Primary

 2: Secondary

 3: Higher
 wealth_index (SPSS Code: V190): This variable indicates the wealth index of the woman's
household.

 1: Poorest

 2: Poorer

 3: Middle

 4: Rich

 5: Richest

 parity (SPSS Code: V201): This variable represents the number of live births the woman has
had, categorized as parity.

 0: 0

 1: 1–2

 2: 3–4

 3: 5+

 age_at_marriage (SPSS Code: V511): This variable represents the woman's age at the time
of marriage.

 0: 15–19

 1: 20–29

 2: 30–39

 3: 40+

 husband_education_level (SPSS Code: V704): This variable represents the highest level of
education attained by the woman's husband.

 0: No education

 1: Primary

 2: Secondary

 3: Higher
 working_status (SPSS Code: V714): This variable indicates whether the woman is currently
working.

 0: No

 1: Yes

 Muslim: This is a derived binary variable indicating if the woman's religion is Muslim
(SPSS Code: V130).

 0: Not Muslim

 1: Muslim

 Sikh: This is a derived binary variable indicating if the woman's religion is Sikh (SPSS
Code: V130).

 0: Not Sikh

 1: Sikh

 Christian: This is a derived binary variable indicating if the woman's religion is Christian
(SPSS Code: V130).

 0: Not Christian

 1: Christian

 Professionals: This is a derived binary variable indicating if the husband's occupation falls
under the Professional category (SPSS Code: V701).

 0: Not Professional

 1: Professional

 Managers: This is a derived binary variable indicating if the husband's occupation falls
under the Managerial category (SPSS Code: V701).

 0: Not Managerial

 1: Managerial

 Sales\Service: This is a derived binary variable indicating if the husband's occupation falls
under the Sales/Service category (SPSS Code: V701).

 0: Not Sales/Service
 1: Sales/Service

 Agricultural\Industrial: This is a derived binary variable indicating if the husband's

occupation falls under the Agricultural/Industrial category (SPSS Code: V701).

 0: Not Agricultural/Industrial

 1: Agricultural/Industrial

 more_earning: This is a derived binary variable indicating if the woman earns more than her
husband (SPSS Code: V746).

 0: Woman does not earn more

 1: Woman earns more

 dont_know: This is a derived binary variable indicating if the woman doesn't know about the
earning comparison (SPSS Code: V746).

 0: Knows about earning comparison

 1: Don't know

 ipv: This is the dependent variable, indicating the presence of intimate partner violence.

 0: No experience of IPV

 1: Experience of IPV

Additionally, the analysis considers different types of domestic violence as outcome variables in
the descriptive statistics: Emotional Violence, Less Severe Physical Violence, Severe Physical
Violence, and Sexual Violence.

Data and Methodology

The National Family Health Survey 2019-21 (NFHS-5), the fifth in the NFHS series, provides
information on population, health, and nutrition for India, each state/union territory (UT), and for
707 districts as on March 31st 2017. The first round of the NFHS was conducted in 1997–98 all
over the country and the fourth round was conducted in 2019-21. Because of the growing
research on DV, which concluded a negative effect of DV against women‘s health, it is
important to include information on DV in the NFHS. All five NFHS surveys have been
conducted under the stewardship of the Ministry of Health and Family Welfare (MoHFW),
Government of India. MoHFW designated the International Institute for Population Sciences
(IIPS), Mumbai, as the nodal agency for all the rounds of NFHS. Funding for NFHS-5 was
provided by the MoHFW, Government of India. ICF, USA provided technical assistance through
the Demographic and Health Surveys (DHS) Program, which is funded by USAID. Assistance
for the Dried Blood Sample (DBS) component of the survey was provided by the Indian Council
of Medical Research (ICMR) and the National AIDS Research Institute (NARI), Pune. NFHS-5
fieldwork for India was conducted in two phases— Phase-I from 17 June 2019 to 30 January
2020 covering 17 states and 5 UTs and Phase-II from 2 January 2020 to 30 April 2021 covering
11 states and 3 UTs — by 17 Field Agencies and gathered information from 636,699 households,
724,115 women, and 101,839 men.

Data for this study were obtained from the NFHS-4 (2015–2016) coordinated by the IIPS. The National
Family Health Survey (NFHS) 5, points to rising instances of domestic and sexual violence against
women in the state. The survey shows that married women, between the ages of 18-49, who have ever
experienced spousal violence, has more than doubled from 20.6 in 2014-15 to 44.5%. NFHS-5 fieldwork
for Karnataka was conducted from 10 July, 2019 to 11 December, 2019 by Nielsen India where data was
gathered from 26,574 households, 30,455 women, and 4,516 men. The survey shows that married women,
between the ages of 18-49, who have ever experienced spousal violence, has more than doubled from
20.6 in 2014-15 to 44.5%.In our study we have total sample of 9255, obtained from women who have
been interviewed and have responded.

Dependent Variable

The present study considered DV as dependent variable (D.V.). In NFHS-4, information was
obtained from never-married women on their experience of violence committed by anyone and
from ever-married women on their experience of violence committed by their current and former
husbands and by others, more specifically, violence committed by the current husband for
currently married women and by the most recent husband for formerly married women. For the
present study, we have considered only ever-married women. It also provides information on
four types of violence against women: sexual, emotional, and severe and less severe physical
violence. The violence was measured by asking all ever-married women if their husbands ever
committed the following to them:

Less severe physical violence: Pushing, shaking, throwing something to the woman, slapping,
punching or hitting by a harmful object, kicking or dragging, strangling or burning, threatening
with a knife/gun/any other weapon.

Emotional violence: Ever been humiliated by husband/partner, ever been threatened with harm
by husband/partner, ever been insulted or made to feel bad by husband/partner.

Sexual violence: Ever been physically forced into unwanted sex by husband/partner, ever been
forced into other unwanted sexual acts by husband/ partner, ever had arm twisted or hair pulled
by husband/partner, ever been physically forced to perform sexual acts respondent did not want
to.
Severe physical violence: Bruises, eye injuries, dislocations, severe burns, wounds, broken
bones, broken teeth, and others.

Independent Variables

The present analysis focuses on three sets of independent variables: (a) household-level factors
(wealth index, marital status, husband drinks); (b) individual-level factors (age of women,
women‘s education, working status of women); and (c) community level factors (types of
residence and religion).

Figure 1. Framework for the determinants of domestic violence.

Statistical Analysis
To examine the association of socioeconomic variables and DV, a dichotomous variable for DV
has been generated. DV is considered ―yes‖ (coded-1) if ever experienced any form of violence
and ―no‖ (coded-0) if never experienced any form of violence. The association between DV and
all socioeconomic variables which are thought to be associated with DV is examined using the
bivariate analysis which produces chi-square p values for the significance of the measure of
association. All covariates that are significant at 5% (p < .05) are considered potential covariates
in the logistic regression analysis. Binary logistic regression models were fitted to estimate the
relationship between sociodemographic variables and violence against married women.
Furthermore, the impact of various background characteristics on women‘s experience of DV
has been assessed through a multiple logistic regression analysis. Logistic regression models are
fitted wherein the model has three forms of violence combined viz. physical, emotional, and
sexual violence as one D.V. The probabilities are interpreted in terms of odds ratios (OR) which
gives the likelihood of experiencing the event in the different categories of a covariate as
compared to a reference category. All statistical analyses were conducted using Python in a
Jupyter Notebook environment.

Cleaning and Renaming Columns:

Initially, the dataset contained numerous coded variables (e.g., ‗V701‘, ‗D105A‘) that were not
intuitive for interpretation. These columns were renamed to more descriptive labels, such as
‗husband_occupation‘, ‗alcoholic_partner‘, and ‗women_earn_more‘, improving the clarity of
subsequent analysis. Irrelevant or redundant variables were removed to streamline the dataset
and reduce computational complexity.

Handling Missing Values:

A detailed missing value analysis was carried out. The variable ‗V746‘ (indicating whether
women earned more than their partners) had the highest proportion of missing values at 97.7%.
Other variables with over 88% missing values included ‗D105D‘, ‗D103A‘, ‗D105J‘, ‗D105I‘,
‗D105H‘, ‗D105E‘, ‗D113‘, ‗D105C‘, ‗D103B‘, ‗D105B‘, ‗D105K‘, ‗D103C‘, ‗D105A‘,
‗V701‘, and ‗V704‘. Variables like ‗V714‘ (partner‘s education level) and ‗V511‘ (age at first
marriage) also had substantial missingness at 84.9% and 26.5%, respectively. To manage this,
variables with excessive missingness were excluded from the analysis. For ‗V511‘, missing
values were imputed using the mode, while for others with smaller proportions of missingness,
rows were dropped. This careful handling of missing data was essential to ensure unbiased and
valid regression results.

Creating Composite Violence Variables:

To measure intimate partner violence comprehensively, a binary IPV (Intimate Partner Violence)
variable was constructed. This variable combined indicators of emotional, sexual, and both less
severe and severe physical violence. If a respondent reported experiencing at least one form of
these types of violence, the composite IPV variable was coded as 1 (violence experienced);
otherwise, it was coded as 0. This holistic measure enabled a broader and more inclusive analysis
of domestic violence patterns.

Recoding and Binning Variables:

Several variables were recoded for better interpretability. Age was grouped into categories (e.g.,
15–24, 25–34), and education levels were binned into ‗No education‘, ‗Primary‘, ‗Secondary‘,
and ‗Higher‘. Similarly, the wealth index, parity, working status, and husband‘s occupation were
categorized meaningfully. These transformations helped ensure statistical stability and made the
model results easier to interpret.
Running Chi-square Tests:
Prior to regression, bivariate chi-square tests were performed to explore associations between
categorical socioeconomic factors and different types of violence. These tests helped identify
statistically significant variables and guided the selection of covariates for multivariate modeling.

Summarizing Socioeconomic Distributions:

Descriptive statistics were calculated to understand the demographic and socioeconomic
composition of the sample. Variables such as residence type, wealth quintile, religion, and
educational attainment were summarized using frequency tables and cross-tabulations, providing
essential context for the regression models.

One-Hot Encoding for Modeling:

To incorporate categorical variables into the logistic regression models, one-hot encoding was
applied. This converted variables like religion, occupation, and residence into binary dummy
variables, preventing incorrect ordinal assumptions and ensuring that the model accurately
reflected the categorical nature of these variables.

Addressing Class Imbalance through Resampling

An initial inspection of the IPV variable revealed an imbalanced distribution, with 4,265 women
not reporting IPV and 2,213 reporting IPV. To mitigate bias in the predictive modeling process,
the dataset was resampled to balance the two classes. After resampling, both IPV and non-IPV
categories had 4,265 observations. This step was critical to avoid skewed model estimates
favoring the majority class and allowed for more reliable and generalizable findings. Although
the exact technique used (e.g., oversampling the minority class or undersampling the majority)
was not specified, balancing class distributions is a standard practice in binary classification
tasks like logistic regression.

Logistic Regression Analysis

Given that the dependent variable, IPV, is binary, logistic regression was chosen as the primary
modeling approach. It models the log-odds of experiencing IPV as a function of multiple
socioeconomic and demographic predictors.

Three primary regression approaches were employed:

1. Logistic Regression with Balanced Data:

This model was estimated on the resampled, balanced dataset to ensure that the model
was not biased toward the majority class. It included variables such as age, residence,
working status, education, wealth, parity, marital status, age at marriage, and husband-
related factors (education, occupation), along with religion and women‘s relative earning
status.
 2. L2 Regularized Logistic Regression:
To prevent over fitting particularly in the presence of many predictors and potential
multicollinearity. An L2-regularized logistic regression was applied. This model included
a penalty term controlled by an alpha value (α = 0.1), which helped stabilize the estimates
and reduce model variance.
3. Regression Modeling of IPV:
Following data preparation, a logistic regression model was estimated to identify
significant predictors of IPV. The dependent variable was the binary IPV indicator, while
independent variables included a comprehensive set of socioeconomic, demographic, and
behavioral factors (e.g., alcohol use by partner, wealth, education, parity, working status,
age at marriage). The model also incorporated encoded religion and occupation
categories. Odds ratios and confidence intervals were calculated to interpret the strength
and direction of associations between the predictors and the likelihood of IPV. Across all
models, the binary IPV indicator served as the dependent variable, while the
independent variables were drawn from the full set of cleaned and recoded
socioeconomic factors. The models aimed to estimate the effect of each predictor—
controlling for the others—on the odds of a woman experiencing intimate partner
violence.

Variables and Categories in the Model

A conceptual framework for predicting the socioeconomic determinants of DV in Karnataka
India is proposed in Figure 1. This framework suggests that DV is a function of individual-,
household-, and community-level variables.

Individual-level variables: The individual-level variables used in the analysis are age of women
(subsequently categorized into age groups), education of women (illiterate, primary, secondary,
higher), and working status.

Household-level variables: The household-level variables included in the analysis are wealth of
the family (poorest, poorer, middle, richer, richest), marital status (married, widowed, divorced,
separated) and husband‘s alcohol use (yes/no).

Community-level variables: The community-level variables included in the analysis are type of
place of residence (rural or urban), religion (Hindu, Muslim, Christian, no religion, and others).

Results
For the current research, a dataset was gathered from a total of 2737 married women within the
age range of 15 to 49 years. These participants were drawn from both rural and urban areas of
Karnataka, India. Among these women, 1067 individuals (29.3%) fell within the age bracket of
15 to 29 years, while 1968 individuals (70.7%) were aged between 30 and 49 years. In terms of
religious affiliation, a significant portion of the participants, specifically 2634 individuals (87%),
identified as Hindus, whereas 335 individuals (11%) identified as Muslims. In the context of
socioeconomic classification, the distribution was as follows: 33.7% of the participants were
categorized as belonging to the middle class, 19.9% were from a poorer socioeconomic class,
and 25.4% were classified as part of the richer socioeconomic stratum.

Table.1. Prevalence of DV among women in the socioeconomic characteristics.

Alcoholic Partner No. of Women %

No 5884 63.6%
Yes 3371 36.4%

Age Group %
15–24 / Younger 2256 24.4%
25–49 / Older 6999 75.6%

Residence %
Urban 1889 20.4%
Rural 7366 79.6%

Highest Education No. of Women %

(Women)
No education 3271 35.3%
Primary 1459 15.8%
Secondary 3597 38.9%
Higher 928 10.0%

Working Status (Women) No. of Women %

Not working 1751 18.9%
Working 7504 81.1%

Age at Marriage No. of Women %

Below 18 5684 61.4%
18 and above 3531 38.2%
Missing / Other 40 0.4%

Religion No. of Women %

Hindu 6687 72.3%
Muslim 719 7.8%
Christian 1148 12.4%
Sikh 283 3.1%
Buddhist 143 1.5%
Jain 7 0.1%
Other 1 0.0%
No Religion 2 0.0%
Don‘t Know 5 0.1%
Missing / Unclassified 260 2.8%

Wealth Index No. of Women %

Poorest 2993 32.3%
Poorer 2362 25.5%
Middle 1631 17.6%
Rich 1171 12.7%
Richest 1098 11.9%

Husband Education Level No. of Women %

No education 2066 22.3%
Primary 1609 17.4%
Secondary 4455 48.1%
Higher 1096 11.8%
Missing 29 0.3%

Husband Occupation No. of Women %

Group
Not working / Unemployed 517 5.6%
Professional / Managerial 528 5.7%
Clerical / Sales / Services 194 2.1%
Skilled Manual 1585 17.1%
Unskilled Manual 6431 69.5%

Women Earn More Than No. of Women %

Partner
Yes 1742 18.8%
Same 4912 53.1%
Partner earns more 1941 21.0%
Don‘t know 504 5.4%
Missing 156 1.7%

Parity (Number of No. of Women %

Children)
None 524 5.7%
One child 6312 68.2%
Two children 1875 20.3%
Three children 416 4.5%
Four or more 128 1.4%
Fig 1: Prevalence of domestic violence
Type of Violence N %
Less severe Physical
Violence Pushing, shaking or throwing things 582 21.3
Twisting arm or pulling hair 1025 37.4
Slapping 421 15.4
Punching 527 19.3

Severe Physical Violence

Choking or burning
152 5.6
Threatening, attacking with a knife, gun or any
115 4.2
other weapon
Kicking, dragging or beating 492 18

Sexual Violence Ever been forced into other unwanted sexual acts 119 4.3
by husband/partner

Ever been physically forced to perform sexual acts 183 6.7

respondent didn't want to

Ever been physically forced into unwanted sex 253 9.2

by husband/partner

Emotional Violence Ever been humiliated by husband/partner 511 18.7

Ever been threatened with harm by 394 14.4

husband/partner

Ever been insulted or made to feel bad by 480 17.5

husband/partner

Table 2: Distribution of study subjects based on prevalence of types of violence‘s

Fig.2: Pie chart showing percentage prevalence of all the types Domestic violence

Variance Inflation Factor (VIF) Analysis

feature VIF

const 0.000000
alcoholic_partner 1.066392

age_group 1.171605

residence 1.323497

highest_education 1.890697

wealth_index 1.989703

parity 1.267533

age_at_marriage 1.153802

husband_education_level 1.569808

working_status 1.026058

Muslim 1.057802

Sikh 1.082072

Christian 1.088265
 Others 1.015009

Professionals 2.020695

Managers 1.366152

Sales\Service 3.427787

Agricultural\Industrial 4.210874

marital_status 0.000000

more_earning 1.027453

dont_know 1.030267

The VIF analysis generally shows values below 5, suggesting that multicollinearity is not a
severe issue among most of the independent variables. However, 'Agricultural\Industrial' and
'Sales\Service' have slightly higher VIF values, and 'sexual_violence',
'less_severe_physical_violence', and 'severe_physical_violence' (when included as predictors)
also show elevated VIF, which is expected given the nature of these violence variables and their
potential overlap with the overall IPV outcome.
Table 3. Percent Distribution of Sampled Women According to Background Characteristics

Socioecono Em Em Em Le Le Le Se Se Se Se Se Se
mic otio otio otio ss ss ss ve ve ve xu xu xu
Characteri nal nal nal Se Se Se re re re al al al
stic Vio Vio Vio ve ve ve Ph Ph Ph Vi Vi Vi
len len len re re re ysi ysi ysi ole ole ole
ce ce ce Ph Ph Ph cal cal cal nc nc nc
No Yes p- ysi ysi ysi Vi Vi Vi e e e
(%) (%) val cal cal cal ole ole ole No Ye p-
ue Vi Vi Vi nc nc nc (% s val
ole ole ole e e e ) (% ue
nc nc nc No Ye p- )
e e e (% s val
No Ye p- ) (% ue
(% s val )
) (% ue
)

age_group 86. 13. 1.0 70. 29. 0.4 91. 8.4 0.5 92. 7.2 0.9
2 8 000 2 8 16 6 84 8 93
2 5 4

working_sta 86. 13. 0.0 69. 30. 0.2 91. 8.7 0.6 93. 6.6 0.0
tus 8 2 275 5 5 88 3 45 4 08
2 4 4

highest_edu 87. 12. 0.0 72. 27. 0.0 92. 7.5 0.0 93. 6.5 0.0
cation 4 6 000 6 4 00 5 00 5 00
0 0 0

wealth_inde 87. 12. 0.0 72. 27. 0.0 92. 7.4 0.0 93. 6.7 0.0
x 1 9 000 8 2 00 6 00 3 00
0 0 1
husband_oc 86. 13. 0.0 73. 26. 0.0 93. 6.5 0.0 93. 6.8 0.4
cupation 9 1 559 4 6 00 5 07 2 03
0 2 5

women_ear 85. 14. 0.0 70. 29. 0.0 90. 9.3 0.0 91. 8.2 0.0
n_more 9 1 002 1 9 00 7 00 8 00
0 0 1

alcoholic_p 84. 15. 0.0 66. 33. 0.0 90. 9.9 0.0 91. 8.2 0.0
artner 8 2 000 8 2 00 1 00 8 00
0 0 0

residence 87. 12. 0.0 72. 28. 0.0 92. 7.5 0.0 93. 6.5 0.0
2 8 001 0 0 00 5 00 5 00
0 0 4

religion 92. 8.0 0.0 68. 31. 0.0 89. 10. 0.0 93. 6.1 0.0
0 000 6 4 00 8 2 00 9 03
0 3 8

parity 84. 16. 0.0 67. 32. 0.0 90. 9.9 0.0 92. 7.4 0.0
0 0 000 3 7 00 1 00 6 66
0 0 2

marital_stat 86. 13. 1.0 70. 30. 1.0 91. 8.5 1.0 92. 7.2 1.0
us 2 8 000 0 0 00 5 00 8 00
0 0 0

husband_ed 87. 12. 0.0 70. 29. 0.0 92. 7.4 0.0 93. 6.4 0.0
ucation_lev 3 7 000 2 8 00 6 00 6 00
el 0 0 0
age_at_mar 86. 14. 0.0 74. 25. 0.0 92. 7.9 0.0 92. 8.0 0.3
riage 0 0 136 1 9 00 1 00 0 79
0 0 1

The table reveals the prevalence of different socioeconomic characteristics and the percentage of
women within each category who reported experiencing different forms of violence. For
instance, women with alcoholic partners reported a higher percentage of experiencing all forms
of violence compared to those without alcoholic partners. Similarly, women with no education or
lower wealth index also showed a higher prevalence of violence. The p-values indicate the
statistical significance of the association between each characteristic and the type of violence.

The Logistic Regression Models

First, bivariate logistic regression models are fitted by considering the socioeconomic variables,
which are significant at the chi-square test separately, and the results are presented in Table 3.
The OR is estimated along with the p values given within the braces. The aim of these beginning
models is to see whether there is any variable which may not be important to consider as a
potential covariate for DV in the multiple model. All p values are significant at 5% level in the
column for sexual physical violence except the two variables (residence and working status).
Residence is also found to be insignificant for Emotional violence which in turn indicates that all
other variables are potential to influence DV.

In order to assess overall effect of various variables on Domestic violence, the Multiple Logistic
Regression model was applied (Table 4). The multiple model (table 4) is fitted by considering
DV as a combined effect of all forms of violence viz. emotional, sexual, less severe and severe
physical violence. The variables that were statistically significant after performing stepwise
regression analysis were included in the multiple logistic regression model. Table 4 shows that
on including different parameters one by one, R2 change is significant, thus table 4 has five
socioeconomic characteristics that are significant and included in our model. A statistically
significant association was observed between woman‘s working status, women‘s educational
level such as graduation & primary school education, drinking habits of husband, type of
residence and wealth index with domestic violence may be considered as independent risk
factors for developing domestic violence. The result of fitting IPV models is presented in Table 4
which consists of estimated OR and corresponding p values.

Table 3. Results of Bivariate Logistic Regression Analysis Between Socioeconomic

Characteristics and Prevalence of Various Forms of Domestic Violence.
Model I: Emotional Violence
Variable OR 2.5% CI 97.5% CI p-value
alcoholic_partner 1.206 1.039 1.4 0.0138
residence 0.888 0.72 1.097 0.2706
highest_education 1.077 0.983 1.179 0.1097
wealth_index 1.012 0.941 1.088 0.7439
parity 1.066 0.961 1.181 0.2268
age_at_marriage 1.08 0.93 1.254 0.3125
husband_education_level 0.912 0.838 0.994 0.0349
working_status 1.236 1.028 1.486 0.0241
less_severe_physical_violence 7.928 6.719 9.355 <0.0001
severe_physical_violence 2.715 2.255 3.27 <0.0001
sexual_violence 4.081 3.343 4.982 <0.0001
Muslim 1.583 1.23 2.037 0.0004
Sikh 0.932 0.731 1.189 0.5724
Christian 1.028 0.631 1.674 0.913
Others 1.245 0.702 2.207 0.4529
more_earning 1.108 0.925 1.326 0.2672
dont_know 1.027 0.787 1.34 0.8446

Model II: Severe Physical Violence

Variable OR 2.5% CI 97.5% CI p-value
alcoholic_partner 2.882 2.614 3.177 <0.0001
residence 1.031 0.897 1.184 0.6704
highest_education 0.891 0.839 0.947 0.0002
wealth_index 0.888 0.846 0.932 <0.0001
parity 1.135 1.058 1.217 0.0004
age_at_marriage 0.838 0.758 0.927 0.0006
husband_education_level 0.999 0.946 1.055 0.975
Muslim 1.273 1.067 1.518 0.0073
Sikh 0.471 0.399 0.555 <0.0001
Christian 0.476 0.337 0.673 <0.0001
Others 0.489 0.322 0.744 0.0008
more_earning 0.955 0.845 1.08 0.4617
dont_know 1.264 1.058 1.511 0.0099
Professionals 0.573 0.407 0.808 0.0015
Managers 1.034 0.688 1.553 0.8721
Sales/Service 0.905 0.712 1.151 0.4165
Agricultural/Industrial 1.152 0.93 1.427 0.1942

Model III: Less Severe Physical Violence

Variable OR 2.5% CI 97.5% CI p-value
alcoholic_partner 3.337 2.841 3.92 <0.0001
residence 0.862 0.677 1.097 0.2267
highest_education 0.982 0.889 1.085 0.7219
wealth_index 0.867 0.799 0.941 0.0006
parity 1.175 1.059 1.303 0.0023
age_at_marriage 0.809 0.686 0.955 0.0122
husband_education_level 0.857 0.783 0.939 0.0009
Muslim 1.944 1.488 2.541 <0.0001
Sikh 0.636 0.486 0.833 0.001
Christian 0.508 0.263 0.984 0.0447
Others 1.32 0.778 2.241 0.3038
more_earning 1.421 1.182 1.707 0.0002
dont_know 1.636 1.254 2.135 0.0003
Professionals 0.765 0.437 1.341 0.3493
Managers 1.07 0.554 2.067 0.8397
Sales/Service 0.77 0.526 1.127 0.1784
Agricultural/Industrial 0.913 0.66 1.264 0.5826

Model IV: Summary (Final)

Variable OR 2.5% CI 97.5% CI p-value
const 0.185971 0.0 inf 1.000000e+00
alcoholic_partner 3.231531 2.855592 3.656963 3.991176e-77
residence 0.719335 0.648811 0.797524 3.915398e-10
highest_education 0.922707 0.826858 1.029668 1.505711e-01
husband_education_level 0.927441 0.832576 1.033115 1.712457e-01
working_status 1.440104 1.101124 1.883439 7.734223e-03
Muslim 2.465047 1.866103 3.25623 2.117127e-10
Sikh 0.953274 0.740447 1.227276 7.104720e-01
Christian 0.61281 0.336809 1.114982 1.088102e-01
Others 0.918128 0.491741 1.714233 7.885984e-01
more_earning 1.491798 1.236487 1.799826 2.960991e-05
dont_know 1.665351 1.255588 2.208842 4.009728e-04

Table.4: Adjusted Logistic Regression Analysis Between Socioeconomic Characteristics and

Prevalence of Various Forms of Domestic Violence.

Logistic Regression Results (with balanced data)

Variable B Exp(B) 95% CI Sig.

alcoholic_partner 1.0 2.7 (2.5 - 3.0) < 0.001

age_group -0.0 1.0 (0.9 - 1.1) 0.4

residence 0.0 1.0 (0.9 - 1.1) 1.0

highest_education -0.1 0.9 (0.8 - 0.9) < 0.001

wealth_index -0.1 0.9 (0.8 - 0.9) < 0.001

parity 0.1 1.2 (1.1 - 1.2) < 0.001

age_at_marriage -0.2 0.8 (0.8 - 0.9) < 0.001

husband_education_level 0.0 1.0 (1.0 - 1.1) 0.6

working_status 0.0 1.0 (0.9 - 1.2) 0.6

Muslim 0.3 1.3 (1.1 - 1.5) < 0.001

Sikh -0.5 0.6 (0.5 - 0.7) < 0.001

Christian -0.7 0.5 (0.4 - 0.7) < 0.001

 Professionals -0.5 0.6 (0.4 - 0.8) < 0.001

Managers 0.3 1.4 (1.0 - 2.0) 0.1

Sales\Service -0.0 1.0 (0.8 - 1.2) 0.8

Agricultural\Industrial 0.2 1.2 (1.0 - 1.5) 0.1

more_earning -0.1 0.9 (0.8 - 1.1) 0.4

dont_know 0.3 1.3 (1.1 - 1.6) < 0.001

const -0.2 0.8 (0.6 - 1.1) 0.2

B = slope of the gradient in logarithmic scale, Exp (B) = antilog of B = Adjusted Odds ratio

Confusion Matrix:

[[1195 634]

[ 380 568]]

Accuracy : 0.6

Precision: 0.5

Recall : 0.6

F1 Score : 0.5
AUC-ROC : 0.7

The confusion matrix shows the model's performance in classifying instances of IPV. The
accuracy of 0.6 indicates that the model correctly classified 60% of the instances. The
precision, recall, and F1 score are all at 0.5. The AUC-ROC value of 0.7 suggests a reasonable
ability of the model to distinguish between the two classes.

The table of regression results provides the coefficients (B), odds ratios (Exp(B)), 95%
confidence intervals, and p-values for each predictor. For example, the odds ratio of 2.7 for
'alcoholic_partner' indicates that women with alcoholic partners have 2.7 times higher odds of
experiencing domestic violence, and this effect is statistically significant (p < 0.001).
Similarly, higher education (highest_education) and higher wealth index (wealth_index) are
associated with slightly lower odds of experiencing domestic violence.

Discussion

The results of the logistic regression analysis provide several important insights into the
socioeconomic determinants of domestic violence. Across the various models, the presence of
an alcoholic partner consistently emerges as a strong predictor, indicating that women with
alcoholic partners have significantly higher odds of experiencing domestic violence. This
finding aligns with extensive existing literature that identifies alcohol abuse as a major risk
factor for IPV. The odds ratios, ranging from 2.7 to 3.3 across different models, underscore
the substantial impact of this factor.

Higher levels of education (highest_education) and a higher wealth index (wealth_index)

generally show a trend towards lower odds of experiencing domestic violence, although the
significance and magnitude of these effects vary across the models. The balanced logistic
regression model suggests a statistically significant but modest protective effect of both higher
education and higher wealth. These findings are consistent with previous research indicating
that socioeconomic empowerment, through education and financial stability, can reduce
women's vulnerability to violence.

The variable 'parity' (number of children) shows a statistically significant positive association
with the odds of domestic violence in some models, suggesting that women with more
children might be at a higher risk. This could be related to increased household stress or other
complex dynamics that warrant further investigation.

Age at marriage (age_at_marriage) appears to be associated with lower odds of domestic

violence in some models, particularly the standard logistic regression. This could indicate that
women who marry at a later age might have greater autonomy or resources, although this
finding requires careful interpretation and might be influenced by other factors.

Religion also emerges as a significant predictor in several models. Muslim women tend to
have higher odds of experiencing domestic violence compared to the reference group, while
Sikh and Christian women show lower odds. These findings highlight the potential influence
of religious and cultural norms on the prevalence of IPV, consistent with existing research.

The results for husband's education level (husband_education_level) are not consistently
significant across the models. In some standard logistic regression models, lower husband's
education is associated with higher odds of domestic violence, aligning with previous studies.
However, this effect is not pronounced in all analyses.

Working status (working_status) shows a statistically significant positive association with the
odds of domestic violence in one of the standard logistic regression models. This finding,
which suggests that working women might be at a higher risk, is complex and has been
reported in some prior research, potentially due to challenges to traditional gender roles or
increased exposure to risk factors.

The variables related to husband's occupation (Professionals, Managers, Sales\Service,

Agricultural\Industrial) show varying levels of significance and effect across the models,
suggesting that specific occupational categories might be associated with different odds of
domestic violence, but these associations are not consistently robust in the provided analyses.
The variables 'more_earning' (woman earns more) and 'dont_know' (don't know about earning
comparison) are statistically significant in some models. 'more_earning' shows a positive
association with the odds of domestic violence, supporting the idea that women who earn
more than their partners might face a higher risk, possibly due to shifts in traditional power
dynamics. The positive association for 'dont_know' is interesting and might reflect underlying
communication issues or power imbalances within the relationship.

The comparison of results across the different models (balanced, regularized, standard) reveals
both consistencies and some variations in the significance and magnitude of the effects. The
strong positive association of 'alcoholic_partner' is consistent across all models. The protective
effect of higher education and wealth index is generally observed, although its significance
varies. The inclusion of L2 regularization in one model tends to yield slightly higher odds
ratios for some significant predictors.

The VIF analysis indicates that multicollinearity is generally not a major concern, although
some of the violence-related variables show higher VIF, which is expected due to their close
relationship. The relatively low VIF values for most socioeconomic predictors suggest that
their effects on domestic violence can be interpreted with reasonable confidence.

Overall, the findings highlight the complex interplay of various socioeconomic factors in
influencing the likelihood of domestic violence. The strong and consistent effect of an
alcoholic partner underscores the critical need to address alcohol abuse in efforts to prevent
IPV. The trends observed for education, wealth, and religion also align with broader societal
patterns and existing research, suggesting that empowering women through education and
economic stability, and addressing cultural and religious norms, are important aspects of
violence prevention strategies.

Conclusion

This dissertation has explored the socioeconomic determinants of domestic violence through a
statistical analysis of a comprehensive dataset. The findings from multiple logistic regression
models reveal several key factors that are significantly associated with the likelihood of
experiencing intimate partner violence. The most prominent and consistent predictor across
the analyses is the presence of an alcoholic partner, which dramatically increases the odds of
domestic violence. This underscores the urgent need for interventions targeting alcohol abuse
as a crucial component of domestic violence prevention.

The analysis also suggests that higher levels of education and a higher wealth index tend to
have a protective effect against domestic violence, although the strength and significance of
this effect vary across the models. These findings reinforce the importance of women's
socioeconomic empowerment through education and financial stability as a strategy for
reducing their vulnerability to abuse.
Furthermore, the study identifies religion as a significant factor, with Muslim women showing
higher odds and Sikh and Christian women showing lower odds of experiencing domestic
violence compared to the reference group. This highlights the potential influence of cultural
and religious contexts on the prevalence of IPV and suggests the need for culturally sensitive
approaches to prevention and intervention.

The findings related to parity, age at marriage, working status, and women earning more than
their partners also provide valuable insights into the complex dynamics of domestic violence,
indicating that factors related to household structure, marital timing, and economic roles can
influence the risk of abuse.

In conclusion, this research contributes to a deeper understanding of the socioeconomic factors

that shape the risk of domestic violence. The identification of key predictors like an alcoholic
partner and the trends observed for education, wealth, and religion provide valuable
information for informing targeted interventions and policy initiatives aimed at reducing the
prevalence of domestic violence and promoting the safety and well-being of women.

Limitations of the Study

This study is subject to certain limitations that should be considered when interpreting the
findings. The cross-sectional nature of the dataset, while providing a snapshot of associations
at a particular point in time, does not allow for the establishment of causal relationships
between the socioeconomic factors and domestic violence. It is possible that the relationship is
bidirectional or influenced by other unobserved factors.

The reliance on self-reported data might also introduce biases, as respondents may underreport
or misreport their experiences due to social stigma, fear, or other reasons. The high percentage
of missing values in several variables, particularly those related to specific experiences and
husband's characteristics, could also impact the representativeness of the data and the
generalizability of the findings. The specific methods used to handle these missing values
were not detailed in the provided information, which further adds to this limitation.

The definition and measurement of domestic violence can also vary, and the current analysis
relies on the specific measures available in the dataset. Different measures might yield
different results. Additionally, the study focuses on a specific set of socioeconomic variables,
and other potentially important factors that were not included in the dataset could also
influence the risk of domestic violence.

Finally, while the resampling technique was used to address class imbalance, it is important to
note that such techniques can sometimes lead to overfitting or other issues that might affect
the generalizability of the model's performance.
Recommendations

Based on the findings of this study and its limitations, several recommendations can be made
for future research and policy initiatives.

Future research should aim to utilize longitudinal data to explore the causal relationships
between the identified socioeconomic factors and domestic violence. This would involve
following individuals over time to observe how changes in their socioeconomic circumstances
are associated with changes in their experience of violence. Qualitative research could also
provide deeper insights into the lived experiences of women facing domestic violence and the
complex interplay of socioeconomic factors in their lives.

Further investigation into the role of religion and cultural norms in shaping attitudes towards
and experiences of domestic violence is warranted. Culturally sensitive research approaches
could help to understand the specific mechanisms through which religious and cultural factors
influence the risk of IPV.

Given the strong association between alcoholic partners and domestic violence, targeted
interventions addressing alcohol abuse are crucial. These interventions should include
awareness campaigns, accessible treatment programs for individuals struggling with alcohol
dependence, and support services for women living with alcoholic partners.

Policy initiatives should continue to focus on women's socioeconomic empowerment through

education, skills training, and employment opportunities. Efforts to reduce gender disparities
in education and the labor market can contribute to enhancing women's autonomy and
reducing their vulnerability to violence.

Strengthening support systems and resources for victims of domestic violence, particularly in
rural areas and among marginalized communities, is essential. This includes ensuring access
to legal aid, counseling services, safe shelters, and economic assistance.

Finally, future research should explore the impact of specific interventions and policies aimed
at preventing domestic violence, taking into account the complex interplay of socioeconomic
factors identified in this and other studies. Evaluating the effectiveness of different approaches
in various socioeconomic and cultural contexts is crucial for developing evidence-based
strategies to address this pervasive social problem.

References

 IPV exists along a continuum from a single episode of violence to ongoing

battering. https://pmc.ncbi.nlm.nih.gov/articles/PMC4525403/ Globally, 30%–
38% of all women who have been in a relationship have experienced physical
and/or sexual violence by their intimate partner.
 Results: The weighted prevalence of DV against women in India in 2019-2021
was 31.2%. https://www.cureus.com/articles/277862-prevalence-and-
predictors-of-domestic-violence-in-india-complex-sample-analysis-of-a-
nationally-representative-study-conducted-between-2019-and-2021
Approximately 28.5%, 13.1%, and 5.7% of women reported experiences of
physical, emotional, and sexual violence, respectively.
 In India, 32% of ever-married women reported having experienced physical,
sexual, or emotional violence by their husbands in their lifetime.
https://pmc.ncbi.nlm.nih.gov/articles/PMC10403108/ The most common type
of spousal violence, in India, is physical (28%), followed by emotional (14%),
and sexual (6%).
 Intimate partner violence against women is a major violation of their human
rights and a public health issue (Heise, 1993; Ellsberg et al., 2001), and it has
been acknowledged by the World Health.
https://www.elibrary.imf.org/view/journals/001/2024/239/article-A001-en.xml
Organization as a "Global Hidden Pandemic"(WHO, 2021).
 As per some records, approximately 30 per cent of married women in India
between 18 and 49 face domestic violence. https://www.business-
standard.com/india-news/nearly-30-of-married-indian-women-face-domestic-
violence-shows-data-123051400486_1.html
 According to the National Family Health Survey (NFHS), 2019-2021, ―29.3
per cent of married Indian women between the ages of 18 and 49 have
experienced domestic/sexual violence; 3.1 per cent of pregnant women aged 18
to 49 have suffered physical violence during their pregnancy.‖
https://www.business-standard.com/india-news/nearly-30-of-married-indian-
women-face-domestic-violence-shows-data-123051400486_1.html
 Global estimates published by the World Health Organisation (WHO) indicate
that ―about 1 in 3 (35%) of women worldwide have experienced either physical
and/or sexual intimate partner violence or non-partner sexual violence in their
lifetime.‖
https://www.google.com/search?q=https://www.swayam.info/resources/violenc
e-facts-figures/
 World Health Organization (WHO) has recognized IPV as a ―global hidden
epidemic‖. https://pmc.ncbi.nlm.nih.gov/articles/PMC11193235/ Worldwide,
one-third of the women have experienced IPV.
 IPV affect general health and reproductive health of women, causing chronic
pain, injuries, fractures, disabilities, unwanted pregnancy and over expose to
contraceptive pills, increasing vulnerability to sexually transmitted diseases.
https://pmc.ncbi.nlm.nih.gov/articles/PMC2253590/
 Violence against women also results in substantial macroeconomic and
household income losses.
https://www.elibrary.imf.org/view/journals/001/2024/239/article-A001-en.xml
Economic growth suffers from less hours worked (absenteeism) and reduced
productivity (presenteeism) (Duvvery and others 2013) which impairs earnings
for individuals and households (United Nations 2005).
 The NFHS-5 (2019-21) India report revealed that approximately 32% of
 women aged 15-49 who had ever been married experienced physical,
emotional, or sexual intimate partner violence (IPV) in the 12 months
preceding the survey. https://pmc.ncbi.nlm.nih.gov/articles/PMC10522783/
 The NFHS-5 data for the urban section shows Karnataka with the highest
(drastic increase from 20.6% to 44.5 %) percentage and Kerala with the lowest
rate (reflecting reversal or lowering from previous data) of spousal
violence.(https://www.google.com/search?q=https://vc.bridgew.edu/cgi/viewco
ntent.cgi%3Farticle%3D3062%26context%3Djiws%23:~:text%3DThe%2520
NFHS%252D5%2520data%2520for,previous%2520data)%2520of%2520spous
al%2520violence.)
 Karnataka was the worst affected state, with 47.3% of women facing DV.
https://pmc.ncbi.nlm.nih.gov/articles/PMC11370985/
 Karnataka was the worst affected state, with 47.3% of women facing DV.
https://www.cureus.com/articles/277862-prevalence-and-predictors-of-
domestic-violence-in-india-complex-sample-analysis-of-a-nationally-
representative-study-conducted-between-2019-and-2021
 In Karnataka, 44% of married women surveyed in the year 2019-2021 claimed
they had faced domestic violence.
https://www.hindustantimes.com/cities/bengaluru-news/nhfs5-survey-data-
karnataka-ranks-no-1-in-domestic-violence-cases-101652076272244.html
 The NFHSs use two-stage sampling techniques for rural areas and three-stage
sampling techniques for urban areas. https://brieflands.com/articles/semj-
148693 Sample selections in rural areas were made in two stages based on PPS
villages, which were designated as primary sampling units (PSUs) for the
random selection of households.
 As identified by feminist theory, patriarchy is the root cause of domestic
violence, whereby males keep women subordinate sometimes with the use of
violence (Martin 1976; Yllo & Strauss 1990).
https://vc.bridgew.edu/cgi/viewcontent.cgi?article=3062&amp;context=jiws
 Patriarchal societies are characterized by deeply entrenched gender norms and
expectations that dictate women's roles and responsibilities.
https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0318350
These norms emphasize duties such as properly preparing food and childcare,
seeking permission before leaving the house, obeying the husband and in-laws,
and fulfilling marital sexual obligations. Transgressing from these pre-
established gender roles often perpetuates PIPV in such patriarchal societies
[19–23].
 Education plays a crucial role, as women with no education exhibit higher rates
of violence. https://www.cureus.com/articles/277862-prevalence-and-
predictors-of-domestic-violence-in-india-complex-sample-analysis-of-a-
nationally-representative-study-conducted-between-2019-and-2021
 Men with less than 5 years of education were more likely to abuse their wives
and were living in poor
conditions.(https://www.researchgate.net/publication/350440822_Socioeconom
ic_Determinants_of_Domestic_Violence_in_Northeast_India_Evidence_From
_the_National_Family_Health_Survey_NFHS-4)
 We confirm previous claims that violence is less common if women and men
are well educated; we also note that acceptance of domestic violence appears to
be related to the respondent's education level.
https://www.tandfonline.com/doi/full/10.1080/15325024.2023.2259292
 Our analysis showed that when husbands are more educated than their wives,
women are more likely to experience certain types of IPV.
https://pmc.ncbi.nlm.nih.gov/articles/PMC11663272/ Specifically, these
women faced higher odds of less severe physical violence and sexual violence,
but there was no significant link with severe physical violence or emotional
violence.
 "Interpretation of this data needs to be done very sensitively," warned Preet
Rustagi, a junior fellow at the New Delhi-based Center for Women's
Development Studies. https://womensenews.org/2003/11/india-domestic-
violence-rises-education/ "Education is an empowering tool for women and
should not be seen as impacting negatively. In fact, this correlation points to the
imperative need for an attitudinal change among men and society in general."
 Results show a robust statistically significant positive association between
income inequality and IPV in India.
https://www.econstor.eu/bitstream/10419/148026/1/872003299.pdf
 The study concludes that the relationship between wealth and IPV differs
considerably among the included countries and that the risk of IPV is not
necessarily higher among women in lower wealth brackets.
https://pubmed.ncbi.nlm.nih.gov/34011189/
 The result shows that women from rural areas experience higher domestic
violence than women from urban
areas.(https://www.researchgate.net/publication/379529419_The_Rural-
Urban_Gap_in_Domestic_Violence_and_Women's_Economic_Empowerment)
Women from rural areas are also found to be less empowered than women from
urban areas. The study further shows that women's economic empowerment is
significant in reducing domestic violence.
 The HBM proposes that when women have more resources, potential, and
income-generating activities, they can avert violence against them as they are
in a position to bargain (Borraz & Munyo,
2020).(https://www.researchgate.net/publication/389561610_Domestic_Violen
ce_and_Women_Empowerment_across_Religion_in_India_A_study_on_influe
nce_of_various_Determinants) Therefore, when HBM is used to explain
violence against women, increased economic opportunities for women are
associated with a decreased likelihood of domestic violence.
 Women's empowerment through employment, asset ownership, and decision-
making is associated with reduced justification and prevalence of physical
violence. https://pmc.ncbi.nlm.nih.gov/articles/PMC11694290/
 In multivariable analyses of cross-sectional data from seven rural and urban
sites across India, the odds of reported violence were two times greater among
employed women whose husbands were unemployed in comparison to
unemployed women whose husbands worked (OR=2.2, 95% CI: 1.3-3.4).
https://pmc.ncbi.nlm.nih.gov/articles/PMC2791993/
 Women who were unemployed at one visit and began employment by the next
visit had an 80% higher odds of violence, as compared to women who
maintained their unemployed status.
https://pmc.ncbi.nlm.nih.gov/articles/PMC2791993/
 Women whose husbands are unemployed have been found to have an elevated
risk of violence. https://pmc.ncbi.nlm.nih.gov/articles/PMC2791993/
 The interplay between spousal employment status and relative income appears
to be a salient factor influencing domestic violence, potentially linked to
prevailing gender roles and power dynamics within the
household.(https://www.researchgate.net/publication/389561610_Domestic_Vi
olence_and_Women_Empowerment_across_Religion_in_India_A_study_on_i
nfluence_of_various_Determinants)
 The paper primarily focuses on examining the influence of various
determinants across religion, with the following findings guiding its selection
as a key aspect of the
study.(https://www.researchgate.net/publication/389561610_Domestic_Violenc
e_and_Women_Empowerment_across_Religion_in_India_A_study_on_influen
ce_of_various_Determinants) Religion is culturally significant, shaping
societies' socio- economic and political dynamics (Stump, 2008).
 People follow their religious instructions may be with slightly varied manners
in all important events in their
life.(https://www.researchgate.net/publication/389561610_Domestic_Violence
_and_Women_Empowerment_across_Religion_in_India_A_study_on_influenc
e_of_various_Determinants) So, the study of religious influence on various
social parameters are of great relevance.