Scorching heat and shrinking horizons: The impact of rising temperatures on marriages and migration in rural India

Mukherjee, Manisha; Martorano, Bruno; Siegel, Melissa Working Paper Scorching heat and shrinking horizons: The impact of rising temperatures on marriages and migration in rural India UNU-MERIT Working Papers, No. 2024-011 Provided in Cooperation with: Maastricht Economic and Social Research Institute on Innovation and Technology (UNU-MERIT), United Nations University (UNU) Suggested Citation: Mukherjee, Manisha; Martorano, Bruno; Siegel, Melissa (2024) : Scorching heat and shrinking horizons: The impact of rising temperatures on marriages and migration in rural India, UNU-MERIT Working Papers, No. 2024-011, United Nations University (UNU), Maastricht Economic and Social Research Institute on Innovation and Technology (UNU-MERIT), Maastricht This Version is available at: https://hdl.handle.net/10419/326907 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-sa/4.0/ #2024-011 Scorching heat and shrinking horizons: The impact of rising temperatures on marriages and migration in rural India Manisha Mukherjee, Bruno Martorano and Melissa Siegel Published 22 May 2024 Maastricht Economic and social Research institute on Innovation and Technology (UNU-MERIT) email: [email protected] | website: http://www.merit.unu.edu Boschstraat 24, 6211 AX Maastricht, The Netherlands Tel: (31) (43) 388 44 00 UNU-MERIT Working Papers ISSN 1871-9872 Maastricht Economic and social Research Institute on Innovation and Technology UNU-MERIT UNU-MERIT Working Papers intend to disseminate preliminary results of research carried out at UNU-MERIT to stimulate discussion on the issues raised. Scorching Heat and Shrinking Horizons: The Impact of Rising Temperatures on Marriages and Migration in Rural India Manisha Mukherjee∗† Bruno Martorano∗ Melissa Siegel∗ Abstract This study delves into the gendered impacts of rising temperatures from climate change on long-term female migration patterns in rural India. We utilize a panel fixed-effects model to investigate the relationship between temperature levels and migration trends by employing a district-level panel dataset that combines Census data for years 1991, 2001, and 2011, multiple rounds of household surveys, and local weather measurements. Our results showcase statistically significant declining effects of rising temperatures on female rural-rural and rural-urban migration in the country. Specifically, the findings suggest that a 1 ◦C temperature increase is associated with a 22% decline in rural-urban and a 13% decline in rural-rural female migration in an average district in India. This decline is primarily due to diminishing marriage-related migration among women in the northern districts of the country, which have a historically high prevalence of dowry. We identify decreasing agricultural yields from rising temperatures as an underlying mechanism that is reducing resources to finance dowry in northern India. We further note that the declines in female migration are driven by districts with poor access to credit, and more so in the northern districts. JEL Classification: J12; O15; Q54 Keywords: Climate change; Migration; Marriages; Dowry; India ∗Maastricht University and United Nations University MERIT, Maastricht 6211AX, The Netherlands †Corresponding Author. Email address: m[email protected]u.edu. We are grateful to Sonja Fransen for valuable comments. We thank Kunal Sen for helpful feedback and Sisir Debnath at Women in Economy Workshop 2024, ISI Delhi, for detailed comments. We are also thankful to the participants and discussants at the RES Annual Conference 2024, Women in Economy Workshop 2024, ISI Delhi, Environment and Inequality Symposium 2024, Sciences Po, PhD Workshop at 2023 Summer School in Development Economics, Prato, 2023 UNU-MERIT Internal Conference, 2023 UNU Migration Network Symposium, and 2023 MACIMIDE Conference for useful comments. Manisha Mukherjee is grateful to Olexiy Kyrychenko for sharing the shapefiles, Praachi Kumar for help with QGIS, and Shubham Sharma for providing important insights on the initial drafts of the paper. 1 I Introduction Anthropogenic climate change is one of the most pressing issues of our time, with far-reaching impacts on human society. According to the sixth assessment report of the Intergovernmental Panel on Climate Change (IPCC), global warming of 1.1 degrees Celsius above pre-industrial levels is already profoundly affecting human systems (IPCC, 2021). These effects are felt most acutely in underdeveloped countries located in low-latitude regions and near the equator (Castells-Quintana et al., 2017; Cline, 2007; Dell et al., 2012). In this regard, Carleton and Hsiang (2016) argue that climate change, by variably affecting different sub-populations like women and the poor, can lead to substantial changes in demographic structures in the long run. These effects can potentially result in further distortion of existing social and economic inequalities in developing countries. One such demographic effect of climate change is on migration patterns, which has garnered significant attention in recent studies (Adger et al., 2015; Black et al., 2011; Cattaneo et al., 2019; Hoffmann et al., 2021). However, while the literature has extensively discussed the impacts of climate change on migration rates in the global south, there continues to be limited evidence on the gendered impacts on migration. This is critical, given the prevalence of various genderrelated norms surrounding women’s migration in developing countries in South Asia and Africa (Bau, 2021; Khalil & Mookerjee, 2018). In light of this, in this study, we investigate the longrun impacts of climate change on internal female migration in India and explore the mediating role of marriage in shaping those impacts. To do this, we create a unique database combining information on intra-district rural-urban and rural-rural female migration for the years 1991, 2001, and 2011 from the Census of India and district-level data on temperature and precipitation from ERA5 reanalysis weather data (Berrisford et al., 2011). Additionally, we gather multiple waves of household-level surveys by the National Sample Survey (NSS) of India between 1987 and 2011 to further delve into the dynamics. The gendered effects of climate change on migration are quite significant. We find that a 1◦C increase in mean decadal temperature is associated with a 22% decline in female rural-urban migration and a 13% decline in female rural-rural migration in an average Indian district. Our results are also robust to multiple sensitivity tests. They remain robust to the inclusion of state-year trends and when the spatial correlation between the districts is taken into account using Conley standard errors (Conley, 1999). The findings are critical as India has a patrilocal structure where most women migrate to their in-laws’ houses after marriage (Chatterjee & Desai, 2020). In fact, most migrants in India are women who migrate because of marriage and not for economic reasons, which is also due to the low participation of women in labor markets (Afridi et al., 2018; Mehrotra & Parida, 2017). Moreover, the event of marriage in India includes the traditional custom of dowry, in which the bride’s household transfers wealth and resources to the groom’s household during marriage. Despite legal procedures, the custom continues to be highly pervasive in the country(Chiplunkar & Weaver, 2023) and plays a decisive role in a woman’s marriage. In this study, we showcase that the declining effects on female migration are driven by reductions in female marriage migration, primarily in the northern part of the country where the dowry custom has been historically more rooted (Mitchell & Soni, 2021). We find that falling agricultural yields due to rising temperatures (Taraz, 2018) that are eroding agricultural incomes act as an underlying mechanism driving our results. Moreover, we also document that poor access to credit in rural areas is moderating 2 the impacts of increasing temperatures on female migration. Besides, although we can only examine impacts on intra-district female migration due to data limitations, multiple sets of evidence indicate that most marriages occurred within the short distance of a woman’s native place and within the same district during the time period of our study (Chiplunkar & Weaver, 2023; Fulford, 2013)1, which upholds the findings of our study. The results of our study highlight a concerning trend as for many women in India, moving to the city through marriage is a means to broaden their horizons and improve their living standards (Kalpagam, 2008). Climate change, by interacting with traditional gender norm, is insidiously shrinking the prospects of women’s movement to urban areas through marriage, trapping women in rural regions with potential negative consequences on their welfare. Our study makes a significant contribution to multiple strands of literature. We supplement the literature on the demographic effects of climate change (Carleton & Hsiang, 2016). We provide evidence of how climate change is distorting demographic structure in the long run by disproportionately affecting the female sub-population in rural India. We also advance the growing literature on the effects of climate change on migration patterns in developing countries (Cattaneo et al., 2019; Hoffmann et al., 2021). Specifically, our study is related to works by Gray and Mueller (2012) and(Findley, 1994). Gray and Mueller (2012) examine the consequences of droughts on migration in rural Ethiopia, and note that drought events are associated with declines in female out-migration as households have fewer resources to bear wedding expenses as traditional norms dictate that the bride’s household are responsible for spending on wedding expenses. Findley (1994) documents similar observations for Mali, where the occurrence of droughts is related to decreases in migration by women as resources to arrange weddings shrink. In our study, we show that rising temperature levels are associated with falling female marriagerelated marriage migration in rural India. In this way, we also add to the literature on the broader economic implications of the increasing temperature levels in rural India. These studies include shrinking rural consumption levels (Aggarwal, 2020; Liu et al., 2023; Sedova et al., 2020), rising rural mortality rates (Burgess et al., 2014), and intensification of agricultural activities as demand for non-agricultural products shrivels due to falling agricultural incomes (Liu et al., 2023). Moreover, our study augments the literature on how traditional gender-related norms shape gender outcomes in developing countries. For instance, Ashraf et al. (2019) document that in Indonesia and Zambia, groups that practice the custom of bride price tend to have more educated daughters. Lowes and Nunn (2017) note no associations between larger bride prices and earlier marriages and high fertility but even assert that bride price is linked to better quality marriages and high self-reported happiness from the wife. In contrast, in India, studies have recorded perverse gender outcomes due to dowry. For example, Bhalotra et al. (2020) detect that higher dowry costs, reflected in the high cost of gold, result in increased son preference behavior amongst the parents. Sekhri and Storeygard (2014) observe that dry rainfall shocks are linked to increases in the incidence of domestic violence against women and dowry deaths in India2. This is because dowry payments are used to smooth consumption in the event of dry shocks by the groom’s households. Moreover, Calvi and Keskar (2021) conclude that women 1This is discussed further with evidence in Section II.II. 2According to Sekhri and Storeygard (2014) , in India, a dowry death is legally defined as the death of a woman that occurs within seven years of her marriage and is caused by any burns or other bodily injury that do not occur under normal circumstances. 3 who paid higher dowries in India are less likely to be poor than those who did not pay. They also possess high control over resources in the household and have better bargaining power within the household. Calvi and Keskar (2021) add that gender poverty gaps are much more pronounced in households with no or low dowry payments relative to those with high and moderate dowry payments. In this regard, we also supplement the literature on the interaction of climate change, marriage markets, traditional gender norms, and gender-related economic outcomes. Our findings closely align with those of Corno et al. (2020) , who show that drought events increase the risk of early marriages in regions in Sub-Saharan Africa, where the bride price custom is common, in which the groom’s household pays the bride during the event of marriage. They further show that droughts decrease the odds of early marriages in India as resources to finance dowry diminish. In congruence with our results, the declines in their study are driven by regions in northern India, where dowry custom is ubiquitous. Another related study by Trinh and Zhang (2021) contrasts Vietnam, which has a tradition of bride price, and India and records similar outcomes that rainfall shocks soar the likelihood of child marriage in Vietnam but have an opposite effect in India. The rest of this article is organized as follows. Section II discusses the study context, and III provides information on data sources and the methodology. Section IV discusses results, and Section Vdelineates the underlying mechanism and contextual factor shaping our results. Section VI presents the discussion and policy implications. Lastly, Section VII provides the concluding remarks. II Study Context In this section, we provide some background on the agricultural sector and climate change, female marriage migration, and the dowry custom in India. II.I Agricultural Sector and Climate Change in India Despite the falling contribution of the agricultural sector to the national Gross Domestic Product (GDP) of India from 30% in 1981 to 16.5% in 2019 (Gulati & Juneja, 2022), the sector remains a crucial component of the economy. It is a major employer of the rural population (Chand et al., 2017) and employed 43% of the country’s workforce and 62% of the rural population in 2019. It is also a principal source of employment for women in rural areas, with 76% of women working in this sector as of 2011-12 (Gulati & Juneja, 2022). Given the strong dependence on the agricultural sector, climate change poses a significant threat to the Indian economy. Studies show a consistent rise in average temperature levels across the country (Kumar et al., 2006; van Oldenborgh et al., 2018), with rainfall trends remaining relatively unchanged on a national scale but exhibiting more region-specific variations (Dash & Hunt, 2007; Kumar et al., 2006; Mall et al., 2006). Multiple studies have highlighted the negative impact of rising temperatures on agricultural yields in India. Taraz (2018) finds that high temperatures are damaging agricultural yields in all districts of the country, particularly the historically colder districts. This is because hotter districts have relatively better accessibility to inter- and intra-crop adaptation methods that aid in adapting to moderate increases in temperature levels. Another study discusses how temperature spikes of 3-6 ◦C in the Indo-Gangetic 4 plains in 2004 led to a decline in wheat production by over 4 million tonnes, also affecting other crops like mustard, peas, onion, and garlic (Kumar et al., 2011). Simulation models-based analyses have underscored the adverse impact of rising temperatures on crop yields in India. A review by Mall et al. (2006) outlines these models’ consensus on the detrimental effects of climate change both on rice and wheat yields in the country. The consequences of rising temperatures on agricultural yields are concerning as means of adaptation remain lacking, suggesting that these losses could be even more severe in the future. Adaptation strategies, such as changing crop patterns, high input delivery, and use efficiency, can help mitigate some of the losses caused by climate change (Kumar et al., 2011). However, there are several barriers to accessing these methods. Some of these methods also require complementary measures, such as investments in supportive infrastructure, to succeed. For instance, modifying crop varieties is successful mostly when complemented by water and soil conservation measures (Ponce, 2020). Access to irrigation remains a challenge, with only 48.7% of India’s cultivated land being irrigated in 2017-18 (Gulati et al., 2020). The average farm size is small, at 1.3 hectares in 2015-16 (Lowder et al., 2016), which further discourages the adoption of new technologies and alternative cropping patterns (Bryan et al., 2014). Studies also indicate significant reductions in groundwater levels in the country, which can threaten agricultural irrigation in the coming years (Zaveri et al., 2016). II.II Female Marriage Migration in India As discussed, the majority of internal migrants in India are women who migrate to stay with their in-laws after marriage (see Figures 2and 3). Along with patrilocal structure, which partly explains these patterns, the practice of village exogamy, in which a woman marries outside her village, is also responsible for high female out-migration in villages (Fulford, 2013). This practice was common in rural northern India but is becoming widespread in other parts of the country as well (Rao & Finnoff, 2015). However, though most marriages occur outside, they are usually within a short distance of the woman’s native place (Fulford, 2013). Based on the Rural Economic Development Survey (REDS) 1999, Chiplunkar and Weaver (2023) document that almost 78.3% of marriages in India occurred within the same district. This trend is also confirmed by the National Sample Survey (NSS) Employment-Unemployment Survey (NSSEUS) of 2007-08, which found that most migration among married working-age women was intra-district, predominantly from one rural area to another (see Figure A.3). However, despite being a common phenomenon, marriage migration and the state of married female migrants remain remarkably understudied in the literature (Fulford, 2013). The most notable study is by Rosenzweig and Stark (1989) , who, based on a small panel of households in a few villages in South India, explore how female marriage migration can serve as a strategy for consumption smoothing in rural households. They delineate that families often marry their daughters into areas with weather patterns uncorrelated to their own, thereby far off from their own village, using this strategy as a form of insurance. These marriage networks enable the transfer of resources from the married daughter’s in-laws to her natal family and vice versa during the years of poor harvest. These findings are challenged by empirical evidence, as discussed above, that shows that most marriages occur within shorter distances, thereby, in areas with weather correlated to the native places. Fulford (2013) further corroborates this by showcasing that transfer of resources is highly uncommon between households connected via marriage links. 5 Moreover, using measures capturing rainfall volatility, Fulford (2013) demonstrates that the fraction of women migrating for marriage is negatively associated with rainfall volatility and adds that parents in areas of high rainfall volatility tend to marry their daughters to closer locations. The marriage of a woman in India typically involves the transfer of dowry, which we explain in detail in the following subsection. II.III Dowry Custom and Preference for Urban Grooms The custom of dowry involves the transfer of wealth from the bride’s household to the groom’s household during the event of the marriage. This practice is highly common in the country, and, although historically rooted amongst upper-caste Hindus in northern India (Dyson & Moore, 1983; Srinivasan & Lee, 2004), recent evidence indicates a widespread adoption of the custom across various castes, religions, and regions (Ifeka, 1989; Srinivasan & Bedi, 2007; Waheed, 2009). Traditionally, dowries comprised gold and silver jewelry, but modern dowries have expanded to include consumer goods like refrigerators and washing machines (Bhalotra et al., 2020). It is a considerable financial burden to households, and a typical dowry can be several times the annual household income (Anderson, 2007; Anukriti et al., 2022). Anukriti et al. (2022) discuss that parents in India start saving for dowry as soon as a daughter is born, and increase their current savings when they expect higher dowries in the future. The fathers of firstborn daughters even have a higher number of work days compared to the fathers of firstborn sons. However, the poorest families in rural areas face various income and behavioral constraints associated with poverty that hinder their ability to save for dowry in advance. Looking at economic explanations of dowry, Becker (1973) regard dowries as pecuniary transfers necessary to clear the marriage market with relatively scarce grooms compared to brides. In contrast, Botticini and Siow (2003) considered it as a feature of virilocal societieswhere married daughters move out of their paternal house, but married sons stay. Nevertheless, as Chiplunkar and Weaver (2023) highlight, the practice became widespread over the years in the countryfrom 38% of marriages involving dowry in the 1920s to 88% by 1975- with dowry amounts tripling from 1945 to 1975. Chiplunkar and Weaver (2023) argue that the expansion in dowry practices and size is linked to changes in groom characteristics. As the average educational level of potential grooms increased in the 20th century, so did the variation in educational attainment. This led to increased competition for the relatively limited number of “high-quality” grooms. In a two-sided matching market, the increase in the number of such grooms with white-collar jobs, who the parents perceived could provide high economic security to their daughters, amplified both the prevalence and size of dowry. Some other studies, such as Anderson (2003) and Calvi and Keskar (2021) , have also highlighted the same. Additional evidence can be found in qualitative and ethnographic studies conducted in villages in India. Studies, such as by Kalpagam (2008) and Chorghade et al. (2006) record the universal practice of transfer of dowry during marriages in villages, and moreover, they find that parents often agree to pay exorbitant amounts of dowry to secure an urban groom. This is primarily because parents in the villages view urban life as less challenging than rural life and prefer for their daughters to marry men with service sector jobs in urban areas, which they believe would ensure a more comfortable and less arduous life for their daughters (Chorghade et al., 2006). These preferences and aspirations of rural parents for urban grooms 6 between 1981-2010. This is done through aggregating yields across the 15 crops, using average district-level crop prices between 1976 and 1980 as weights. Using base-year prices as crop weights removes the impact of climate shocks on prices, as discussed by Taraz (2018) and Duflo and Pande (2007) Furthermore, we examine the moderating role of access to credit at the baseline year as a contextual factor. For this, we focus on district-level access to bank branches in the baseline year of 1980. The data on the number of bank branches in each district in 1980 are obtained from the Basic Statistical Returns (BSR) reports published by the Reserve Bank of India9. Additionally, we source district-level population data from the Census 1981 (Vanneman & Barnes, 2000) and construct a baseline variable of bank branches per capita by dividing the total number of bank branches by the population. Lastly, to explore the changes in dowry values, we utilize the most common data source on dowries in India, the Rural Economic and Demographic Survey (REDS). REDS collected data on economic and demographic variables of rural households across 17 major states in India, which encompass 96% of the total population of the country. It additionally recorded information on the marriages of the household heads, their brothers, sisters, sons, and daughters. We employ the 1999 round of REDS and consider marriages that occurred during the period of our analysis, that is, from 1981 onwards. REDS includes data on the nominal value of the dowry paid from the bride to the groom’s household and the dowry received from the groom to the bride’s household during the marriage. It also has information on the year of marriage. We follow the methodology adopted by Chiplunkar and Weaver (2023) to compute the net real values of dowry paid. We create the net dowry paid variable as the difference between the value of the dowry paid by the bride’s household and the value of the dowry paid by the groom’s household. Thereafter, we convert the net nominal value of dowry paid to the real values using the Wholesale Price Index (WPI)10. III.II Empirical Strategy We conduct a panel fixed-effects estimation, and the main econometric specification is as follows: Yj,t =β0Tj,t +β1Pj,t +ϕj+γt+ϵj,t (2) The variable Yj,t is the outcome of interest in district jand in decade beginning with year t (= 1991,2001,2011). It is the share of rural-urban (rural-rural) intra-district female migrants in the total female population of the district. This is our primary outcome of interest. For comparison and completeness, we also consider the share of rural-urban (rural-rural) intradistrict male/total migrants in the total male/total population of the district. Tj,t represents the average annual temperature in ◦Celsius of district jin the past decade ending in year t.Pj,t is the average annual precipitation in millimetres of district jin the past decade ending in year t. We follow the literature discussing the estimation of impacts of climate change on economic outcomes and include both temperature and precipitation in the specification (Auffhammer et al., 2013; Cattaneo & Peri, 2016) so that we obtain unbiased estimates of the effects of rising temperatures on female migration rates. However, as discussed in Section II.I, following the 9https://dbie.rbi.org.in/DBIE/dbie.rbi?site=publications Accessed 12 May 2024. 10The WPI data for the years 1981 to 1998 is gathered from the Reserve Bank of India. 13 climate change literature on India, which documents consistent rises in average temperatures across the country but region-specific changes in precipitation, we only interpret the coefficients associated with the temperature levels throughout this study. Thus, β0is the main coefficient of interest in this study and captures the effect of rising temperatures on migration rates, and β0<0signifies that migration rates are declining with rising temperatures. Furthermore, we also include fixed effects, ϕj, capturing fixed district characteristics. γt is the year effects that control for changes over time that are consistent across districts. We only include fixed effects as controls to ensure that the model stays parsimonious and to avoid overcontrolling. This is because including other controls, such as socioeconomic characteristics that might be correlated to agricultural productivity, may introduce bias in the estimation due to over-controlling (Cattaneo & Peri, 2016). The standard errors are clustered at the district level and are provided in parentheses in the regression tables. Specification (2) is a reduced-form linear relationship between rural-urban and rural-rural female migration rates and climate variables. In the upcoming sections, we discuss the possible channels through which climate change may shape migration rates, such as changes in agricultural productivity and contextual factors like access to credit. IV Results Table 1illustrates the estimated effects of annual temperatures on intra-district female ruralrural and rural-urban migration rates in India. They indicate a statistically significant negative effect of increasing temperatures on rural-rural and rural-urban female migration rates. Specifically, the results show that a 1 ◦Celsius increase in temperature in an average Indian district is associated with a 22% decrease in the share of female rural-urban migrants and a 13% reduction in the share of female rural-rural migrants. The effects on male migration rates are not statistically significant, aligning with the findings of Liu et al. (2023) . Table 1: Effect of Rising Temperatures on Migration, 1991-2011 Rural-Urban Rural-Rural Total Male Female Total Male Female (1) (2) (3) (4) (5) (6) Temperature −0.004 −0.002 −0.007 −0.018 −0.006 −0.030 (0.003) (0.003) (0.003)** (0.008)** (0.006) (0.012)** Precipitation 0.004 0.003 0.004 −0.001 0.008 −0.010 (0.003) (0.002) (0.003) (0.005) (0.003)** (0.007) District FE Y Y Y Y Y Y Year FE Y Y Y Y Y Y Observations 1,341 1,341 1,341 1,341 1,341 1,341 Note: The dependent variable in (1) is the share of total intra-district rural-urban migrants, in (2) is the share of male intra-district rural-urban migrants, in (3) is the share of female intra-district rural-urban migrants, in (4) is the share of total intra-district rural-rural migrants, in (5) is the share of male intra-district rural-rural migrants, and in (6) is the share of female intra-district rural-rural migrations, in years 1991, 2001, and 2011. The independent variables are the decadal averages of annual temperature (◦C) and annual precipitation (mm). The standard errors are clustered by district in parentheses. *p < 0.10, **p < 0.05, ***p < 0.01 14 We argue that the declines in female migration are primarily due to the decreases in marriagerelated migration. Furthermore, these declines could be driven by the districts in northern India, where the custom of dowry, as previously discussed, has been historically widespread. To check for this, we categorize the districts as northern districts that are located in the northern states11. We exploit this heterogeneity in the historical prevalence of dowry between the northern districts and the rest of India by re-running the previous analysis and allowing the coefficients of temperature and precipitation to vary depending on whether the district is in a northern state. The results are reported in Table 2and match our expectations. In the northern districts, rising temperatures have a pronounced negative effect on female migration rates12. Table 2: Effect of Rising Temperatures on Female Migration in Northern Districts, 1991-2011 Rural-Urban Rural-Rural Female Female (1) (2) Temperature 0.010 0.029 (0.005)** (0.016)* Temperature x North −0.022 −0.075 (0.004)*** (0.012)*** Precipitation 0.006 −0.007 (0.003)*(0.006) Precipitation x North −0.017 -0.040 (0.004)*** (0.015)*** District FE Y Y Year FE Y Y Observations 1,341 1,341 Note: The dependent variable in (1) is the share of female intra-district rural-urban migrants in the years 1991, 2001, and 2011. The dependent variable in (2) is the share of female intra-district rural-rural migrants in the years 1991, 2001, and 2011. The independent variables in (1) and (2) are the decadal averages of annual temperature (◦C) and decadal averages of annual temperature interacted with “North” dummy denoting 1 if the district lies in a major northern state such as Punjab, Uttarakhand, Uttar Pradesh, Haryana, Bihar, Jharkhand, Rajasthan, Madhya Pradesh, and Chhattisgarh, and 0 otherwise. It also includes the decadal averages of annual precipitation (in mm) and decadal averages of annual precipitation interacted with the “North” dummy. The standard errors are clustered by district in parentheses. *p < 0.10, **p < 0.05, ***p < 0.01 It might still be the case that these changes observed through Census data are not due to 11We follow the Government of India’s categorization of the north and north-central cultural zone (Ministry of Culture, Govt. of India, n.d.) to identify the northern states and classify the states Punjab, Haryana, Uttar Pradesh, Rajasthan, Bihar, Rajasthan, Uttarakhand, Chhattisgarh, Jharkhand, and Madhya Pradesh as northern states. We exclude the erstwhile state of Jammu & Kashmir due to the unavailability of Census data in 1991. The state of Himachal Pradesh is also omitted as it has a smaller population and area, and we only consider the major northern states. The state of Uttarakhand is also small but has been considered as it was part of the state of Uttar Pradesh until 2000. Similarly, the states Jharkhand and Chhattisgarh are also considered as they belonged to the states of Bihar and Madhya Pradesh, respectively, until the year 2000. 12In an additional analysis, we allow the temperature coefficients to vary based on whether the district is in southern states of India, where the custom historically has not been prevalent. The results are reported in Appendix C, highlighting no negative effects for southern districts. These results complement the findings of Corno et al. (2020) . 15 changes in marriage-related female migration. Therefore, we further corroborate our analysis by providing more evidence on the marriage channel using the NSS-EUS survey for the years 1987-88, 1999-2000, and 2006-07. These waves of EUS had modules on migration where the survey respondents, that is, the members of the household, were asked if their current place of enumeration was different from their last usual place of residence. If so, the primary reason for this migration is recorded. Thus, using this data and the survey weights provided by NSS, for each of these waves of EUS, we construct an outcome variable that is the share of the total married working-age female migrants in the district who have undergone intra-district rural-urban migration due to marriage in last ten years. We compute similar measures for intra-district rural-rural migration. The approach is aligned with the measurement of migration variables using the Census data, but here, the variables are refined further to capture only the marriage-related migration by women. The computation of climate variables stays the same. We then carry out a panel fixed effects regression in a similar manner as before with district-fixed and year effects. The results are presented in Table 3, and findings from regression (3) indicate a negative association between increases in temperature and female rural-urban marriage migration in the northern regions. These findings lend further support to the initial results, suggesting that the declines in female migration observed previously are driven by decreases in marriage-related migration in northern India. Table 3: Effects of Rising Temperatures on Marriage Migration, NSS Survey, 1987-2007 Rural-Urban Rural-Rural Rural-Urban Rural-Rural Female Female Female Female (1) (2) (3) (4) Temperature 0.009 0.0005 0.030 −0.001 (0.016) (0.044) (0.017)*(0.053) Temperature x North −0.049 −0.007 (0.023)** (0.069) Precipitation 0.001 −0.023 0.022 −0.013 (0.015) (0.033) (0.015) (0.034) Precipitation x North −0.026 −0.070 (0.023) (0.059) District FE Y Y Y Y Year FE Y Y Y Y Observations 881 881 881 881 Note: The dependent variables in (1) and (3) are the share of married female intra-district rural-urban migrants of working age who have migrated due to marriage in NSS rounds 1987-88, 1999-00, and 2007-08, respectively. The dependent variables in (2) and (4) are the share of married female intra-district rural-rural migrants of working age who have migrated due to marriage in NSS rounds 1987-88, 1999-00, and 2007-08, respectively. The independent variables in (1) and (2) are the decadal averages of annual temperature (◦C) and precipitation (in mm). The independent variables in (3) and (4) are the decadal averages of annual temperature (◦C) and decadal averages of annual temperature interacted with “North” dummy denoting 1 if the district lies in a major northern state such as Punjab, Uttarakhand, Uttar Pradesh, Haryana, Bihar, Jharkhand, Rajasthan, Madhya Pradesh, and Chhattisgarh, and 0 otherwise. The regression also controls for the decadal average of precipitation (in mm) and decadal average of precipitation interacted with “North” dummy. The standard errors are clustered by district and are in parentheses. *p < 0.10, **p < 0.05, ***p < 0.01 Moreover, we posit that these declines in marriage migration by women are due to the 16 diminishing agricultural income and wealth from soaring temperatures in rural areas, which, in turn, is dwindling the resources for dowry in northern areas. We provide suggestive evidence of shrinking dowry amounts in rural northern India in figure 4below. The figure displays a negative relationship between net dowry paid by brides’ households and rising temperatures. IV.I Robustness Tests In this section, we discuss the robustness tests we implement to further validate our main results regarding female migration rates in Table 1. These tests include adding state-year trends to the regression equation, re-running the analysis using alternative climate variables, and accounting for spatial autocorrelation. The results of robustness tests are attached in Appendix B. State-year trends. In the first robustness test, we add state-year trends to our regression equation to capture state-specific time trends. The findings, presented in Table A3, show a slight increase in the magnitude of the estimates. Importantly, the direction and statistical significance of the results stay consistent with the initial findings. Alternative climate variables. We explore the sensitivity of our main results to different measures of temperature and precipitation. For this, we aggregate grid-level weather data to district-level outcomes by calculating the weighted average of all grid points within 100 kilometers of each district’s geographic center, using an inverse-square weighting of the distances, as suggested by Taraz (2018) . These results, which are displayed in Table A4, affirm the robustness of the initial findings. The estimates remain statistically significant and consistent with the main results. The magnitudes of the estimated coefficients have increased slightly, indicating that the effects of rising temperature on female migration rates might be even higher than previously estimated. Conley standard errors. In the third robustness test, we adjust the standard errors to consider possible spatial auto-correlation (Conley, 1999; Hsiang, 2010). We follow Sekhri and Storeygard (2014) and set the distance cutoffs at 100, 150, and 200 kilometers. The results, detailed in Table A5, demonstrate that the estimates remain robust when applying Conley standard error adjustments for these distance cutoffs. V Mechanisms and contextual factors In this section, we first investigate the underlying mechanism driving our results. We explain that declines in agricultural yields due to rising temperatures are eroding the wealth of rural households. This, in turn, harms their ability to finance the dowry payments. In addition, we test whether the effects of rising temperatures on female migration rates vary with a contextual factor, such as poor access to credit. V.I Agricultural Yields and Dowries As we discussed before, evidence supported by simulation models and observational data shows the detrimental effects of temperature rises on agricultural yields in India. We replicate these findings for the time period of our study and, specifically, carry out an analysis to dissect the effects for the northern districts. We follow Taraz (2018) and Liu et al. (2023) to measure the yearly aggregate agricultural yield at the district level. As detailed in Section III, we use the VDSA data on agricultural yields and compute the aggregate agricultural yield at the district 17 level for the time period of our analysis, 1981-2010. Thereafter, we conduct a panel fixed-effects estimation with the natural logarithm of the aggregate agricultural yields as the dependent variable13. As expected, the results in Table 4show a marked decline in agricultural yields due to rising temperatures. Table 4: Effects of Rising Temperatures on Agricultural Yields, 1981-2010 Log Agr. Yield (1) Temperature −0.065 (0.019)*** Precipitation 0.044 (0.010)*** District FE Y Year FE Y Observations 8,693 Note: The dependent variables in (1) is the natural logarithm of the aggregate agricultural yields, between the years 1981 and 2010. The independent variables in (1) are the average annual temperature (◦C) and precipitation (in mm). The standard errors are clustered by district and are in parentheses. *p < 0.10, **p < 0.05, ***p < 0.01 How do these declines in agricultural yields translate into lower resources for dowry? The gradual declines in agricultural yields from rising temperatures over the years are decreasing income for agricultural households. However, rural households in India exhibit consumption patterns characterized by habit formation, meaning they slowly adjust their consumption in response to income volatility (Khanal et al., 2019), often delving into their savings to finance consumption. As the agricultural households gradually and disproportionately adjust their consumption to the decreases in agricultural incomes, far fewer resources are left for savings for dowry14. The social and economic costs of not being able to get a daughter married are high (Anukriti et al., 2022), which can compel a few agricultural households to possibly overcome these losses in savings by adapting to rising temperatures and finding ways to save enough for dowry. However, this could be specifically more difficult for relatively poorer households15, who are unable to adapt and fail to save enough. As the cost of an unmarried daughter is high, these households 13We execute the following econometric analysis: lnYj,t =η0Tj,t +η1Pj,t +ϕj+γt+ϵj,t Here, Yj,t is the aggregate agricultural yield in district jand year t.Tj,t is the annual average temperature in ◦C and Pj,t is the annual average precipitation in mm in district jand year t. We control for district fixed effects using ϕjand year effects through γt. 14We cannot completely discount that consumption levels are not changing. Studies have discussed the negative effects of rising temperatures on the consumption of rural households (Liu et al., 2023). More likely, both consumption and savings are drying up. 15Approximately 70% of the rural population in India has zero to little landholdings (Asher et al., 2022). Research shows considerable linkages between the size of landholdings and access to adaptation to climate change (Baland & Robinson, 2008; Bryan et al., 2009; Deressa et al., 2009; Dustmann & Okatenko, 2014). 18 also need to get their daughters married, and they do so by resorting to paying lower amounts of dowry. We provide suggestive evidence of this in Figure 4below. It displays a negative relationship between the net dowry amounts paid by the brides’ households and the increasing temperature levels in the northern rural regions16. With lower amounts of dowry, these households are settling for “low” quality grooms residing in the rural regions rather than “high” quality grooms in urban areas for their daughters. Figure 4: Net Dowry Paid by Brides and Temperature Levels in Northern Districts, 1981-1998 Note: The plot exhibits the real value of the net dowry paid by the bride’s household in rural areas in the northern districts for marriages between 1981 and 1998 in 1981 Rupees in the vertical axis. The nominal values are converted to real values using WPI. On the horizontal axis are the average temperature levels in ◦C in the corresponding marriage year in the district. Each point in the scatter plot represents the surveyed household with a bride in marriage years between 1981 and 1998. Source: Authors’ own illustration using the REDS 1999 and ERA5 data. In this regard, access to credit that can enable these households to borrow for dowries during the marriage years can play a critical role and considerably shape our initial results. We explore this issue in the following subsection. V.II Access to Credit: Heterogeneity Analysis In this subsection, we examine how access to credit in a district at the baseline year can influence the effects of rising temperatures on the female migration rate. Access to credit can moderate the impacts in two broad ways. It can enhance adaptation to climate change by agricultural households by facilitating investment in agricultural technologies or encouraging diversification into the non-agricultural sector17. Moreover, better access to credit can increase the ability to 16This is similar to Chowdhury et al. (2020) , who discuss how income shocks in rural Bangladesh impact marriage payments. A positive income shock increases the bride price and dowry amounts, but a negative income shock reduces them. 17We run similar heterogeneity analysis based on access to irrigation at the baseline in the district, as better access to irrigation can also aid in adaptation to climate change. The results are provided in Appendix Dand do not show any discernible effects. 19 save in advance and enable borrowing during the marriage year to finance dowry and wedding expenses (Anukriti et al., 2022). For this analysis, we identify districts with low access to banks in the baseline year 1980. We determine whether the per capita number of bank branches was below the national median and create a binary variable to represent this condition: it equals 1when the per capita number of bank branches in a district in 1980 is below the national median and 0otherwise. The spatial plot in Figure A.4 in the Appendix shows significant variation in access to bank branches across districts in 1980. To assess how it moderates the effect of rising temperatures on female migration rates, we carry out a variation of the regression equation (2). We allow the coefficients attached to temperature to change based on whether the district had poor access to credit at the baseline. The findings in Table 5outline that limited access to banks significantly shapes the impact of temperature increases on female migration, both rural-rural and rural-urban, with a more substantial effect on the latter. Our results confirm that the access to credit constraints faced by poor agricultural households prevent them from saving for dowry in advance, challenge securing loans during the marriage years, and perhaps even constrain their ability to adapt to rising temperatures. This, in turn, is intensifying the effect of climate change on the marriage prospects of rural women and the marriage-related migration by them. In magnitude terms, these results show that a 1 ◦Celsius increase in temperature is associated with a 25% decline in female rural-urban migration in districts with low access to banks and a 19% decrease in female ruralrural migration. The lower magnitude of the declining effect associated with rural-rural female migration hints that rural households in districts with poor access to credit are perhaps resorting to marrying their daughters to other rural areas rather than to urban areas by offering lower dowry amounts. Additionally, we adjust the analysis to examine the effects specifically in the northern districts. The findings are depicted as results of regressions (3) and (4) in Table 5. Unsurprisingly, we observe that the diminishing effects are driven by northern districts and, more so, by northern districts with limited access to credit at the baseline. 20 Table 5: Effect of Rising Temperatures on Female Migration and Heterogeneous Effects by Access to Banks, 1991-2011 Rural-Urban Rural-Rural Rural-Urban Rural-Rural Female Female Female Female (1) (2) (3) (4) T0.005 −0.019 0.010 0.028 (0.004) (0.013) (0.005)*(0.016)* T x Low Bank Access −0.011 −0.037 (0.003)*** (0.012)*** T x North −0.018 −0.057 (0.005)*** (0.013)*** T x North x Low Bank Access −0.006 −0.034 (0.002)*** (0.013)*** P0.016 −0.016 0.006 −0.006 (0.005)*** (0.007)** (0.003)*(0.006) P x Low Bank Access -0.020 0.008 (0.005)*** (0.011) P x North −0.012 −0.021 (0.004)*** (0.022) P x North x Low Bank Access −0.004 −0.010 (0.004) (0.028) District FE Y Y Y Y Year FE Y Y Y Y Observations 1,326 1,326 1,326 1,326 Note: The dependent variables in (1) and (3) are the share female intra-district rural-urban migrants, respectively, in years 1991, 2001, and 2011. The dependent variables in (2) and (4) are the share female intra-district rural-rural migrants, respectively, in years 1991, 2001, and 2011. The independent variables in (1) and (2) are the decadal averages of annual temperature (◦C), decadal averages of annual temperatureinteracted with “Low Bank Access” dummy denoting 1 if the per capita Bank branches at the baseline was lower than the national median and 0 otherwise, decadal average of annual precipitation (in mm), and decadal average of annual precipitation (in mm) interacted with the “Low Bank Access” dummy. The independent variables in (3) and (4) are the decadal averages of annual temperature (◦C), decadal averages of annual temperature interacted with “North” dummy denoting 1 if the district lies in a major in a major northern state such as Punjab, Uttarakhand, Uttar Pradesh, Haryana, Bihar, Jharkhand, Rajasthan, Madhya Pradesh, and Chattisgarh, and 0 otherwise, and decadal averages of annual temperature interacted with “North” dummy and “Low Bank Access” dummy. The regressions also control for the decadal average of annual precipitation (in mm), and decadal average of annual interacted with the “North” dummy, and decadal average of annual interacted with the “Low Bank Access” dummy. The standard errors are clustered by district in parentheses. *p < 0.10, **p < 0.05, ***p < 0.01 VI Discussion & Policy Implications In this study, we discuss the gender-specific impacts of climate change, particularly the effects of rising temperatures, on long-term female migration patterns in rural India. We provide evidence of temperature increases substantially affecting female rural-rural and rural-urban migration rates. These reductions in female migration rates are mainly driven by declines in marriagerelated female migration in northern districts where the custom of dowry is ubiquitous. The policy implications of this study, particularly concerning the traditional custom of dowry, are multifaceted and complex. Various studies have highlighted the ineffectiveness of the Dowry Prohibition Act of India, with some suggesting legal procedures can inadvertently exacerbate 21 women’s welfare (Calvi & Keskar, 2021,2023). On a different note, Chiplunkar and Weaver (2023) ’s recent study suggests that, as India’s literacy rate is rising sharply, the dowry custom might contract in the future as women’s literacy improves and the supply of educated grooms increases. However, other studies present a contrasting view, arguing that dowry custom would continue to play a fundamental role in marriages in India in the coming years (Beauchamp et al., 2017). This is because grooms increasingly seek educated women as brides but not highly educated. Thus, highly educated women may face challenges in finding suitable grooms and might need to offer substantial dowries for marriage. Nonetheless, an important policy recommendation based on the findings is improving access to credit in rural areas. Burgess et al. (2014) have demonstrated how access to banks can reduce mortality in rural areas during summers in India. Banks facilitate consumption smoothing and provide loans at manageable interest rates, which can be crucial for financing dowry resources and adapting to climate change. For instance, farmers can obtain credit to invest in adaptation strategies such as irrigation, crop diversification, mechanization, and purchasing seeds. Additionally, banking access also offers resources for migration to cities. This has been highlighted by Bryan et al. (2014) , who found that offering small loans to subsistence-based farmers and job matching can reduce their aversion to migration. However, it is critical that financial inclusion policies are gender-sensitive. This ensures that women have equal opportunities to migrate for education or work, in addition to traditional marriage-related migration. Mehrotra and Parida (2017) argue that women’s labor force participation is likely to increase as more women attain education. Yet, this progression depends on women’s mobility and sense of safety when moving out. Therefore, policy interventions are required to first encourage the acceptance of women migrating for reasons other than marriage, such as education or employment opportunities, and make destination places safer. Furthermore, as suggested by Anukriti et al. (2022) , policies that liquidate dowry, such as financial inclusion policies that mitigate barriers to savings and induce savings behavior in poor rural households, can evade distress borrowing at exorbitant rates from informal sources during the time of marriage. Finally, our study highlights the urgent need for state interventions to transform rural economies to be more resilient to climate change. Infrastructure programs, like road construction and electrification that connect rural areas to cities, can enhance access to agricultural markets, reduce information costs, and bolster the uptake of adaptation means by agricultural households. VII Conclusion The literature on climate change and migration in developing countries has expanded significantly in recent years. However, there remains a notable gap in understanding the specific impacts on women, who often constitute a major share of internal migrants due to the patrilocal structure prevalent in these regions. Although female migration appears non-economic, there are often various economic channels underlying the marriage of a woman in these countries. In our study, we demonstrate that rising temperatures are decreasing female migration rates in India, specifically female marriage-related migration in northern regions. We contend that this is due to the historical prevalence of dowry in northern India and identify decreasing agricultural yields from soaring temperatures as an underlying mechanism. In particular, we show that the reductions in agricultural yields are shrinking resources to finance dowry. We further note the 22 Mehrotra, S., & Parida, J. K. (2017). Why is the Labour Force Participation of Women Declining in India? World Development,98, 360–380. https://doi.org/https://doi.org/10.1016/j. worlddev.2017.05.003 Ministry of Culture, Govt. of India. (n.d.). Zonal cultural centers [Accessed: 12-05-2024]. Mitchell, A. M., & Soni, S. (2021). Institutional development and the dowry death curve across states in India. Journal of International Development,33(6), 1026–1042. https://doi. org/https://doi.org/10.1002/jid.3549 https://doi.org/10.1002/jid.3549. Murthi, M., Srinivasan, P. V., & Subramanian, S. (2001). Linking Indian Census with National Sample Survey. Economic and Political Weekly,36, 783–792. https://doi.org/10.2307/ 4410351 Ponce, C. (2020). Intra-seasonal climate variability and crop diversification strategies in the Peruvian Andes: A word of caution on the sustainability of adaptation to climate change. World Development,127, 104740. https://doi.org/https://doi.org/10.1016/j.worlddev. 2019.104740 Rao, S., & Finnoff, K. (2015). Marriage Migration and Inequality in India, 1983–2008. Population and Development Review,41(3), 485–505. https://doi.org/https://doi.org/10.1111/j. 1728-4457.2015.00069.x https://doi.org/10.1111/j.1728-4457.2015.00069.x. Rosenzweig, M. R., & Stark, O. (1989). Consumption Smoothing, Migration, and Marriage: Evidence from Rural India. Journal of Political Economy,97(4), 905–926. https://doi. org/10.1086/261633 Sedlander, E., Talegawkar, S., Ganjoo, R., Ladwa, C., DiPietro, L., Aluc, A., & Rimal, R. N. (2021). How gender norms affect anemia in select villages in rural Odisha, India: A qualitative study. Nutrition,86, 111159. https://doi.org/https://doi.org/10.1016/j.nut. 2021.111159 Sedova, B., Kalkuhl, M., & Mendelsohn, R. (2020). Distributional Impacts of Weather and Climate in Rural India. Economics of Disasters and Climate Change,4(1), 5–44. https: //doi.org/10.1007/s41885-019-00051-1 Sekhri, S., & Storeygard, A. (2014). Dowry deaths: Response to weather variability in India. Journal of Development Economics,111, 212–223. https://doi.org/https://doi.org/10. 1016/j.jdeveco.2014.09.001 Srinivasan, P., & Lee, G. R. (2004). The Dowry System in Northern India: Women’s Attitudes and Social Change. Journal of Marriage and Family,66 (5), 1108–1117. http://www. jstor.org/stable/3600328 Srinivasan, S., & Bedi, A. S. (2007). Domestic Violence and Dowry: Evidence from a South Indian Village. World Development,35(5), 857–880. https://doi.org/https://doi.org/10. 1016/j.worlddev.2006.08.005 29 Taraz, V. (2018). Can farmers adapt to higher temperatures? Evidence from India. World Development,112, 205–219. https://doi.org/https://doi.org/10.1016/j.worlddev.2018.08.006 Trinh, T.-A., & Zhang, Q. (2021). Adverse shocks, household expenditure and child marriage: evidence from India and Vietnam. Empirical Economics,61(3), 1617–1639. https://doi. org/10.1007/s00181-020-01907-2 Vanneman, R., & Barnes, D. (2000). Indian district data, 1961-1991: Machine-readable data file and codebook [College Park, Maryland: Center on Population, Gender, and Social Inequality, http://www.inform.umd.edu/~districts/index.html]. van Oldenborgh, G. J., Philip, S., Kew, S., van Weele, M., Uhe, P., Otto, F., Singh, R., Pai, I., Cullen, H., & AchutaRao, K. (2018). Extreme heat in india and anthropogenic climate change. Natural Hazards and Earth System Sciences,18(1), 365–381. https://doi.org/ 10.5194/nhess-18-365-2018 Waheed, A. (2009). Dowry among Indian Muslims: ideals and practices. Indian Journal of Gender Studies,16(1), 47–75. Zaveri, E., Grogan, D. S., Fisher-Vanden, K., Frolking, S., Lammers, R. B., Wrenn, D. H., Prusevich, A., & Nicholas, R. E. (2016). Invisible water, visible impact: groundwater use and Indian agriculture under climate change. Environmental Research Letters,11 (8), 84005. 30 Appendices A Figures & Tables Figure A.1: Districts in India, 1991 Note: The figure depicts 466 districts in 1991 in India. We excluded the northern union territories of Jammu & Kashmir and Ladakh (erstwhile northern state of Jammu & Kashmir) due to the unavailability of migration data in 1991. We also dropped the union territories Pudducherry and Lakshadweep because of the lack of availability of weather data. The excluded regions are marked in grey. The figure is processed using the district shapefile in QGIS. 31 Figure A.2: Female Migration in India, Census of India, 1991-2011 Note: The figure depicts the average female rural-rural migration in the years 1991, 2001, and 2011 (Left) and average female rural-urban migration in the years 1991, 2001, and 2011 (Right). Source: Authors’ own illustration using Census of India data. The figure is processed using the district shapefile in QGIS. 32 Figure A.3: Female Migration in India, NSS, 1987-2008 Note: (Clock-wise)The first chart shows the share of migrants (Total/men/women) in India in three NSS rounds 1987-88, 1999-2000, and 2007-08. The second plot illustrates the migration of working-age married women by sector. The third plot presents the migration of working-age married women by region. Source: Authors’ own calculations using NSS-EUS data. 33 Figure A.4: Irrigation and Banks per capita at baseline (1980) Note: The figure depicts the distribution of banks per capita (left) and proportion of land irrigated (right) at the baseline year of 1980. The districts are categorized into high banks per capita and low banks per capita based on the national median banks per capita in 1980, which is approximately 4 for every 100,000 people. Similarly, the districts are categorized into high irrigated and low irrigated based on the national median proportion of land irrigated in 1980, which is equal to 0.17. Source: Authors’ own compilation using the Reserve Bank of India (RBI) data and data from ICRISAT (2015) . The maps are processed using the district shapefile in QGIS. 34 Table A1: Temperature and Precipitation, 1981-2010 Year Mean SD Annual Temperature (◦C) 1981-1990 24.0 4.51 1991-2000 24.1 4.44 2001-2010 24.4 4.43 Annual Precipitation (m) 1981-1990 0.00377 0.00251 1991-2000 0.00391 0.00254 2001-2010 0.00393 0.00238 Source: Authors’ own calculation using ERA5 data 35 Table A2: Marriage and gender norms, IHDS, 2011-12 North India South India Rest of India Rural Urban Rural Urban Rural Urban Ask permission to visit health centre 0.87 0.82 0.82 0.84 0.72 0.71 If yes, can you go alone? 0.57 0.71 0.61 0.62 0.70 0.70 Ask permission to visit friend or relative in the neighborhood 0.73 0.73 0.85 0.85 0.60 0.63 If yes, can you go alone? 0.64 0.73 0.63 0.59 0.75 0.74 Ask permission to visit kirana shop 0.42 0.50 0.60 0.62 0.42 0.46 If yes, can you go alone? 0.75 0.84 0.75 0.75 0.81 0.83 Do you practice ghoonghat/ purdah/ pallu? 0.84 0.72 0.11 0.18 0.73 0.50 Family meal eating: men eating first 0.48 0.28 0.14 0.11 0.22 0.14 Does anybody in the family have a bank account? 0.70 0.74 0.65 0.71 0.60 0.75 If yes, is the respondent’s name on the bank account? 0.47 0.55 0.67 0.65 0.46 0.56 Note: The data is gathered from the India Human Development Survey-II, 2011-12. The module with data on ever-married women aged 15-49 is used. The measures are calculated using the survey weights. The large northern states are Punjab, Uttarakhand, Uttar Pradesh, Haryana, Bihar, Jharkhand, Rajasthan, Madhya Pradesh, and Chhattisgarh. The large southern states are Kerala, Tamil Nadu, Karnataka, and Andhra Pradesh. Rest of the states are Jammu & Kashmir, Himachal Pradesh, Sikkim, Arunachal Pradesh, Nagaland, Manipur, Mizoram, Tripura, Meghalaya, Assam, West Bengal, Odisha, Gujarat, Maharashtra, and Goa. Source: Authors’ own calculation using India Human Development Survey, 2011-12 36 B Robustness Tests Table A3: Effect of Rising Temperatures on Migration with State-Year Trends, 1991-2011 Rural-Urban Rural-Rural Female Female (1) (2) Temperature −0.007 −0.033 (0.003)** (0.013)*** Precipitation 0.005 −0.010 (0.003) (0.007) District FE Y Y Year FE Y Y State Year Trends Y Y Observations 1,341 1,341 Note: The dependent variables in (1) and (2) are the share of female intra-district rural-urban and rural-rural migrants, respectively, in the years 1991, 2001, and 2011. The independent variables are the decadal averages of annual temperature (◦C) and decadal averages of annual precipitation (mm). The standard errors are clustered by district in parentheses. *p < 0.10, **p < 0.05, ***p < 0.01 Table A4: Effect of Rising Annual Temperatures on Migration with Alternative Climate Variables, 1991-2011 Rural-Urban Rural-Rural Female Female (1) (2) Temperature −0.008 −0.033 (0.003)*** (0.013)** Precipitation 0.003 −0.010 (0.003) (0.007) District FE Y Y Year FE Y Y Observations 1,341 1,341 Note: The dependent variables in (1) and (2) are the share of female intra-district rural-urban and rural-rural migrants, respectively, in years 1991, 2001, and 2011. The independent variables are the decadal averages of annual temperature (◦C) and decadal averages of annual precipitation (mm). The standard errors are clustered by district in parentheses. *p < 0.10, **p < 0.05, ***p < 0.01 37 Table A5: Effect of Rising Annual Temperatures on Female Migration with Conley Standard Errors, 1991-2011 Variable Standard error Distance cut-off Temperature Precipitation Coefficient SE Coefficient SE Rural-urban Conley standard error 100 −0.007 (0.003)** 0.004 (0.003) Conley standard error 150 −0.007 (0.003)*0.004 (0.004) Conley standard error 200 −0.007 (0.004)*0.004 (0.004) Clustering District −0.007 (0.003)** 0.004 (0.003) Rural-rural Conley standard error 100 −0.031 (0.011)*** −0.010 (0.006)* Conley standard error 150 −0.031 (0.012)** −0.010 (0.006) Conley standard error 200 −0.031 (0.012)** −0.010 (0.007) Clustering District −0.030 (0.012)** −0.010 (0.007) Observations 1,341 1,341 Note: The coefficients are estimated based on the same model as in equation (2). These regressions control for district fixed effects and year effects. The Conley standard errors are computed using the conleyreg() function under the package conleyreg in statistical software R with time lag equal to 0. The distance cut-off are in kilometres. The Conley standard errors are in the parentheses. *p < 0.10, **p < 0.05, ***p < 0.01 C Additional Analysis: Southern States C.I Impacts on Female Migration Rates in the Southern States To check for the effects on female migration rates in the southern states, we run a similar analysis as in equation (2) and allow the temperature coefficients to vary based on whether the district is in the southern states 18. The results are shown in Table A6, and we observe a positive effect of rising temperatures on female migration rates. This positive effect matches with the results documented by Corno et al. (2020) . The authors find that overall, droughts decrease the incidence of early marriages in India as resources to afford dowry reduce following a drought. However, the effects are mainly observed for early-born cohorts in South India by Corno et al. (2020) . As previously discussed, the dowry custom was historically much less prevalent in southern India but is recently becoming common. It could be the case that our model is unable to capture the recent rises in dowry demands in southern India as the data in this study is relatively old. We find suggestive evidence of this when we vary the coefficients based on the year 2011 in Table. The effects on female ruralurban migration in southern India are similar to that in the northern region, albeit statistically significant only at 10%. 18The southern states are identified as those categorized in the southern cultural zone by the Government of India (Ministry of Culture, Govt. of India, n.d.). These states are Karnataka, Andhra Pradesh, Kerala, and Tamil Nadu 38 The UNU-MERIT WORKING Paper Series 2024-01 The green transformation as a new direction for techno-economic development by Rasmus Lema and Carlota Perez 2024-02 Evidence on an endogenous growth model with public R&D by Thomas H.W. Ziesemer 2024-03 Higher Educational Systems and E-resilience by Mindel van de Laar, Julieta Marotta and Lisa de Graaf 2024-04 Estimating the wage premia of refugee immigrants: Lessons from Sweden by Christopher F. Baum, Hans Lööf, Andreas Stephan and Klaus F. Zimmermann 2024-05 Green window of opportunity through global value chains of critical minerals: An empirical test for refining copper and lithium industries By Jorge Valverde- Carbonell and Alejandro Micco A. 2024-06 Migration and Consumption by Roberta Misuraca and Klaus F. Zimmermann 2024-07 The power of narratives: How framing refugee migration impacts attitudes and political support towards refugees by Tobias Hillenbrand, Bruno Martorano, Laura Metzger and Melissa Siegel 2024-08 Strategies of search and patenting under different IPR regimes by Robin Cowan, Nicolas Jonard and Ruth Samson 2024-09 From drains to bridges: The role of internationally mobile PhD students in linking non-mobile with foreign scientists by Rodrigo Ito, Diego Chavarro, Tommaso Ciarli, Robin Cowan and Fabiana Visentin 2024-10 Expanding horizons: The impact of post-doctoral mobility grants on research diversification by Yağmur Yıldız, Diego Chavarro, Fabiana Visentin and Tommaso Ciarli 2024-11 Scorching heat and shrinking horizons: The impact of rising temperatures on marriages and migration in rural India by Manisha Mukherjee, Bruno Martorano and Melissa Siegel