Reviewer: 1 Comments to the Author This paper shows that, in India, areas where drought occurred have more intimate partner violence (IPV), even independent of covariates. The reliance on cross-sectional geographic variation makes this paper hard to interpret. Even a repeated cross-sectional design using district-level changes between NFHS4 and the newly released NFHS5 would have been more informative. Nonetheless, this paper is interesting and important. The paper needs more discussion on correlates and causes of IPV in general, which can then be linked to the discussion of the potential pathways in the specific case of droughts. Perhaps show annualized state and national level precipitation data to give some idea on the trends and fluctuations. And discuss its consequences in India. Add some rationale behind indexing drought one year prior to the IPV report. Are the stressors associated with drought felt afterward and up to a year? Why? If it is unclear when the stressors associated with drought will be felt, consider indexing drought to different periods (perhaps multiple, even simultaneously) before the survey as a sensitivity check. Provide a rationale for specifying drought as you do (ie, below the 30 percentile of precipitation). The definition of draught also seems crude. An alternative approach that gives more variation would be to look at changes in precipitation (eg, mm, percentage, z-score): for example, changes in precipitation from one year to the next, even in interaction with the level of precipitation (mean-centered). Such interaction model would give three main parameters: 1) association of precipitation 12–24 months before survey with IPV (when there has been no change in precipitation); 2) association of change in precipitation between 24–35 and 12–23 months with IPV (for a PSU with average precipitation 12–23 months before survey); and 3) an interaction between the two. Squared terms could be added if there is a concern that too much precipitation would be detrimental. Although the microdata is cross-sectional more could be done with the temporal dimension of the precipitation data. Some covariates identified as confounders (line 36 page 5) may be pathways to some extent (which you point out elsewhere for some of them). Wealth and drinking, for example, could be impacted by drought. Some of them can be dropped but some, like wealth, are most likely also confounders. It’s a limitation that biases estimates downward. I don’t see why it isn’t possible to explore heterogeneity by occupation (line 20 page 9). There are variables that indicate whether they own land for agriculture as well as the size of land and more in the NFHS4. The use of the drought-prone status variable is informative, and the explanations given for the observed heterogeneity reasonable. But this variable may simply reflect geography more broadly. From the simple association observed, many conclusions are drawn, and no attention given to alternative explanations. The language needs to communicate better that much of the interpretations are highly speculative and provide alternative interpretations for the observed associations. Minor comments: Briefly define dowry death (line 44 page 3): eg, “...increased levels of dowry deaths (ie, suicide or murder following marriage related to dowry dissatisfaction) in India.” Page 5 line 10: is it annual precipitation data? Sounds like it is monthly in line 14. Are multilevel models necessary here? Seems like the clustering is a nuisance that can be solved by simply by clustered standard errors. Information on missing data. Reviewer: 2 Comments to the Author -This study combines two different data sets intuitively. The study design is well thought out but the implication of your research and its uniqueness is not coming out very well. -There exists a number of empirical analysis on climate change related events (droughts, extreme rainfall, other natural disasters) and IPV. Please highlight how your work stands out in the existing literature in this domain. What difference does your study makes to what is already known (natural disasters and IPV literature) -Line 32 (Pg. 3) The heading - " Drought in India" better be changed as that section discusses literature on droughts and IPV. -Regarding your precipitation-based and socio-economic drought : The precipitation based drought is at PSU (village/ town level) while socio-economic is at district level and there are huge intra-state variations with huge intra and inter-district variations. How you ensure that these differences in level of analysis are accounted for in your analysis and conclusion? -For your covariate - gender equity there are far better indicators available in your data (NFHS-4), Like attitude of male as well as female partner toward wife beating under certain circumstances, [Pg. 5 line 50] - Whether household depends on agriculture is a critical variable in establishing this relationship. Please check if its possible to include it in your analysis - Since you have data on drought prone and other areas it would be wise to do a differential analysis of drought prone areas as then you can present your work differently and implications would be more clearer. Though you have touched upon this but I feel expanding this would add more value to your work. Drought prone areas are accustomed to extreme conditions and for them it is not a shock, they are better prepared to cope and thus lesser IPV instances -Conclusion focuses on IPV prevention programmes but given the cause of IPV is economic shock to the family and lack of resources the findings have implication for preparedness or programmes for natural disasters, extreme weather events etc. Kindly consider adding the same