Page 77 - IJPS-11-4

International Journal of
Population Studies Droughts and intimate partner violence

Oceania, Latin America, and the Caribbean. There are observations with missing values for covariates and cases
two types of DHSs: the Standard DHS, conducted every in which the value of the variable “d121” (whether the
5 years with a large sample size to allow for time-based respondent had seen her father beat her mother) was 8,
comparison, and the Interim DHS, conducted between the as well as observations with missing values for “s116”
standard DHSs, using a smaller questionnaire and sample (the respondent’s social group). The final sample size
size. The DHSs for India are also referred to as the NFHS. comprises 34,590 observations, with 16,123 from NFHS-4
For consistency, the term NFHS was used in this study. We and 18,467 from NFHS-5.
used pooled data from NFHS-4 (DHS-2015 – 2016) and
NFHS-5 (DHS-2019 – 2021). Data collection for NFHS-4 2.1. Dependent variable
occurred in two phases (January 20, 2015 – December 04, IPV can take various forms. The NFHS measures three
2016), and for NFHS-5, it was conducted in two phases: forms of IPV PV, SV, and EV. A woman is considered
Phase 1 from June 17, 2019 to January 30, 2020, and Phase to have experienced PV if, in the past 12 months, her
2 from January 02, 2020, to April 30, 2021. In total, NFHS-4 intimate partner has pushed, shook, slapped, punched,
interviewed 699,686 women and 112,122 men from kicked, dragged, strangled, or burned her, twisted her arm,
601,509 households, and NFHS-5 interviewed 724,115 or threatened her with a knife. SV occurs when a woman
women and 101,839 men from 636,699 households. is forced into unwanted sexual activities or acts by her
The Couples’ recode file, which contains 63,696 partner. EV includes humiliation, insults, threats of harm,
observations from NFHS-4 and 57,535 observations from or other forms of emotional abuse. Each form of IPV is
NFHS-5, is used in this analysis. Of these, 47,514 women in measured as a dichotomous variable and takes the value of
NFHS-4 and 46,353 women in NFHS-5 were interviewed 1 if the woman has experienced it at any time or in the past
for the domestic violence module. Table S1 summarizes 12 months, and 0 otherwise.
this information. 2.2. Independent variables
The domestic violence module of NFHS-5 was The independent variable in this study explains the effect
administered to women aged 18 – 49 years in a subsample of drought on IPV. The Indian states of Andhra Pradesh,
of households selected for the state module, following Karnataka, and Tamil Nadu experienced severe droughts
a similar structure as that of NFHS-4. According to the in 2018 due to a precipitation deficit from NEM. However,
World Health Organization guidelines on the ethical because data collection in Tamil Nadu was interrupted by
collection of domestic violence data, only one eligible COVID-19, we excluded Tamil Nadu and similar states and
woman per household was randomly selected. The Union Territories. This left Andhra Pradesh and Karnataka
module was not administered if privacy could not be as the two states exposed to NEM drought. We created a
obtained after at least three attempts to ensure respondent binary variable, “Drought,” by coding districts in drought-
confidentiality. affected states as 1 and districts in unaffected states as 0.
For comparability, Dadra and Nagar Haveli and Daman Thus, all districts in Andhra Pradesh and Karnataka are
and Diu were combined in the NFHS-4 sample, as NFHS-5 coded as 1 to proxy for drought exposure from the NEM. In
represents these Union Territories. As NFHS-4 did not impact evaluation terms, Andhra Pradesh and Karnataka
collect data for Ladakh, we excluded this region from represent the treatment group, and the remaining 16 states
NFHS-5. and Union Territories represent the comparison group.
The onset of the COVID-19 pandemic interrupted the 2.3. Control variables
data collection process. Data for 16 states were collected The primary source of our control variables was
in 2019 (Table S2) and completed by February 2020 in covariates commonly recognized in the IPV literature
the Andaman and Nicobar Islands and Lakshadweep. For (Dhanaraj & Mahambare, 2021; Pathak, 2022; Pathak
the remaining states and Union Territories, the data were & Kumar, 2023). We controlled the respondent-level
collected between 2020 and 2021. We used data from 19 characteristics, including current age (a categorical
states and Union Territories, for which the data collection variable with seven categories), education level of
was completed in February 2020. The states and Union both the respondent and her partner (a categorical
Territories included in the analysis are marked with an variable with six categories), age at first cohabitation,
asterisk in Table S2. age at first childbirth, total number of children born,
Due to the legal obligations of the Protection of and respondent’s employment status. In addition, we
Children from Sexual Offenses Act, we included only controlled for whether the respondent had witnessed
respondents aged 18 years or older. We also excluded her father beating her mother.

Volume 11 Issue 4 (2025) 71 https://doi.org/10.36922/ijps.3065