Page 84 - IJPS-11-4

International Journal of
Population Studies Droughts and intimate partner violence

treatment and comparison groups, both before and after The DID estimator reveals an increase in the odds of
exposure to the NEM drought. The results are shown in PV, SV, and EV in the states affected by the NEM drought
Table S3. The following observations were made: across both models. These results are significant at p = 0.01
a. The correlation coefficients between the number of level for all forms of IPV, except for SV in the model with
control issues and the incidence of PV were significant full controls.
at p = 0.01 level across all groups, indicating a strong In the model without controls, exposure to the NEM
association between having a controlling partner and drought is associated with an approximate 83% increase
experiencing PV. in the odds of PV, whereas the effect is slightly reduced
b. The coefficient for the comparison group decreased to a 48% increase when a full set of controls to applied.
from 0.3253 in 2015 – 2016 to 0.2797 in 2019 – 2021, The corresponding increase in odds for SV is 114% in the
suggesting that although the relationship remained model without controls and approximately 25% in the
strong and positive, its magnitude decreased over model with controls, although the estimates for SV are not
time. statistically significant. For EV, the DID estimator indicates
c. Conversely, the coefficient for the treatment group a 103% increase in odds in the model without controls and
increased over the same period, from 0.3727 to 0.429, a 44% increase in the model with a full set of controls.
implying that for respondents in states exposed to These estimates demonstrate that exposure to the NEM
drought, the relation between having a controlling drought is significantly associated with increased odds of
partner and experiencing PV strengthened over time. PV and EV at the 1% significance level.
3.2.4. Controlling behavior of partners and IPV In an auxiliary regression, we calculated the predicted
We estimated the association between the number of probabilities of PV using our DID framework. We
control issues a woman faces and the incidence of PV. As then generated kernel density plots for four groups: the
the direction of the relationship is uncertain, we followed comparison group in 2015 – 2016, the treatment group in
Acock (2018), using the number of control issues reported 2015 – 2016, the comparison group in 2019 – 2021, and the
as the row variable and the incidence of PV as the column treatment group in 2019 – 2021. These plots are presented
variable. We conducted separate analyses for states exposed in Figure 2, and two key observations emerge:
to rainfall shock, using Kruskal’s gamma and Kendall’s tau-b a. The respondents in the treatment group, which
as measures of association to demonstrate concordance. includes the states of Andhra Pradesh and Karnataka,
We hypothesized that a woman facing numerous control had marginally higher predicted probabilities of PV
issues would also report IPV, and conversely, a woman in the baseline (2015 – 2016). This finding aligns with
the results shown in Table 2, where the difference in
reporting IPV would likely report control issues as well.
The results of these analyses are presented in Tables S4-S6. the proportions of respondents facing PV between the
treatment and comparison groups in 2015 – 2016 was
Figure 1A-C show the percentage of women reporting statistically significant at p = 0.10 level.
each form of IPV based on the number of control issues b. The predicted probabilities of PV increased
faced. These figures are derived from the data, as shown in significantly for respondents in the treatment group
Tables S4-S6. Figure 1A shows a clear positive association pre-exposure to the NEM drought. This finding further
between the number of control issues reported and the supports the results presented in Table 2, which shows
incidence of PV. Notably, respondents in non-drought- that the difference in the incidence of PV between the
exposed states reported lower incidences of PV in 2019 comparison and treatment groups has grown nearly
– 2021 when faced with more than two control issues. In eight-fold. In addition, the statistical significance of
contrast, for states exposed to drought, the incidence of this difference was improved to the p = 0.01 level.
PV was higher across nearly every control issue category.
Similarly, Figure 1C shows a strong positive association 3.3.1. Subsample analysis
between controlling behavior and EV, although no clear To further explore the effect of the NEM drought on IPV,
pattern emerges for SV in Figure 1B. we conducted a subsample analysis to examine whether
the impact differs between rural and urban areas. The logic
3.3. Effect of drought on IPV: The DID estimation behind this approach is that if the NEM drought affects IPV
Table 3 presents the results of the DID for both models: through reduced agricultural output and income, its effect
one without controls and one with a full set of controls. would likely be more pronounced in rural areas. The results
Both models incorporated district-level fixed effects, and of this subsample analysis are shown in Table 4. In the rural
standard errors are clustered at the district level. areas of states affected by the NEM drought, the odds of

Volume 11 Issue 4 (2025) 78 https://doi.org/10.36922/ijps.3065