https://orcid.org/0000-0003-1820-5357
![Sample our Global Development journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5929/TOC-Subject-Sample-2021-Global-Development.jpg"})
ABSTRACT
Climate variability has the potential to affect both international and internal migration profoundly. Earlier work finds that higher temperatures reduce agricultural yields, which in turn reduces migration rates in low-income countries, due to liquidity constraints. We test whether access to irrigation modulates this temperature–migration relationship, since irrigation buffers agricultural incomes from high temperatures. We regress measures of international and internal migration on decadal averages of temperature and rainfall, interacted with country-level data on irrigation and income. We find robust evidence that, for poor countries, irrigation access significantly offsets the negative effect of increasing temperatures on internal migration, as proxied by urbanisation rates. Our results demonstrate the importance of considering access to alternative adaptation strategies when analysing the temperature-migration relationship.
Acknowledgments
We thank Julia Bouzaher, François Le Béhot, Ahana Raina, and Emily Zhou for excellent research assistance. We thank Arun Agrawal, Cristina Cattaneo, Simon Halliday, Valerie Mueller, Susan Sayre, Wolfram Schlenker, and Sneha Thapliyal for feedback on this paper. We are grateful for feedback from seminar participants at Smith College, the University of Luxembourg and the University of Neuchâtel as well as conference participants at the 2018 International Workshop on Migration and Environment at ETH Zürich, the 2018 Liberal Arts Colleges Development Economics Conference (LAC-DEV) at Middlebury College, the 2018 Sustainability and Development Conference (SDC) at the University of Michigan, the 2019 Agricultural and Applied Economics Association (AAEA) Annual Meetings, the 5th DIAL Conference on Development Economics, and the 6th FAERE Annual Conference. Katrin Millock is fellow of the French Collaborative Institute on Migration (IC Migrations).
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 We analyse the effect on urbanisation rates as a proxy for rural-urban migration. Despite high natural urban population growth rates, rural-urban migration is the main factor of urbanisation (Jedwab, Christiaensen, and Gidelsky Citation2017).
2 The measure includes area equipped for full/partial irrigation, and areas equipped for spate irrigation, but excludes rainwater harvesting.
3 Negative values are put equal to zero, as in the main specification of Beine and Parsons (Citation2015) and in Cattaneo and Peri (Citation2016), which implies assuming negligible return migration flows and mortality.
4 Daily temperature data would be ideal to explore agricultural channels, as it would allow us to construct daily temperature bins (Schlenker and Roberts Citation2009) or degree days (D'Agostino and Schlenker Citation2016). Unfortunately, widely used daily gridded weather data sets such as ERA-Interim (Dee et al. Citation2011) and the Modern-Era Retrospective Analysis for Research and Applications (Rienecker et al. Citation2011) begin in 1979, corresponding to the modern era of remotely sensed data, and are hence unsuitable to use with our emigration and urbanisation data (which begin in 1960).
5 For countries missing data on maize growing season dates, we instead use average monthly weather based on the entire twelve-month calendar year.
6 Contrary to the urbanisation rates, the emigration rates are skewed and taking the natural logarithm normalises the distribution.
7 Following other authors (Beine and Parsons Citation2015; Cai et al. Citation2016; Cattaneo and Peri Citation2016) we use the 1990 income distribution rather than the initial time period distribution, because of missing values for GDP for earlier years.
8 See Appendix 1.
9 For this exercise we use the mean and standard deviation of 1960's irrigation relative to the set of poor countries: in other words, we use the summary statistics from the third column of Table .
10 We lose 10 countries in the sample for these estimations: Djibouti, Ethiopia, Fiji, Haiti, Sao Tome and Principe, Somalia, Saint Vincent and the Grenadines, Trinidad and Tobago, United Arab Emirates, and Yemen.
11 The list of countries dropped in this specification is the Bahamas, Belize, Bhutan, Botswana, Brunei Darussalam, Cape Verde, Comoros, Cyprus, Djibouti, Equatorial Guinea, Fiji, Gabon, Gambia, Guinea-Bissau, Guyana, Kuwait, Lesotho, Mauritius, Namibia, Oman, Qatar, Saint Vincent and the Grenadines, Sao Tome and Principe, Solomon Islands, Suriname, Swaziland, United Arab Emirates, and Vanuatu.