Climate Change and Migration Project

This repository contains R and python code for my project on climate change and migration in collaboration with Nancy Chau (Cornell University), Amanda D. Rodewald (Cornell University) and Filiz Garip (Princeton University).

This project turned into a research article, titled "How to model the weather-migration link: a machine-learning approach to variable selection in the Mexico-U.S. context", and it was published at the Journal of Ethnic and Migration Studies in 2022 (link to the article).

The code extracts weather data and creates weather measures used for the analyses. It also combines it with survey information from MMP (Mexican Migration Project), and run random-forests models to predict migration. The code for each programming language is stored in separate folders (r_code and python_code). A third folder results stores the results that appeared on the paper (figures and tables).

NOTE: There is a lot of code in this repository! Although I tried to create master files (00-master.R and run.py) that help reproduce the results step by step and added comments to explain what almost all the code does, it may be difficult to navigate all the files. Please reach out if you have a hard time trying to understand some parts of the code structure.

Code

R code

The R code does several operations needed before running the random-forests models for the analysis. In particular:

• It processes Mexico's shapefiles.
• It creates weather information using data from Dayment.
• It computes weather measures described on the paper and it creates files with these variables that will be used with the MMP data using python code.
• It creates Figure 2 on the paper.

Python code

The python code runs random forests using scikit-learn. In particular,

• It manipulates the data containing the MMP survey and Daymet weather measures and prepares the data for the analysis.
• It creates Figures 3, 4, and 5 on the paper.
• It gives the results presented in Tables 2 and 3 on the paper.

Data

Mexican Shapefiles

Data for Mexico's map can be downloaded here: https://www.inegi.org.mx/app/mapas/. This project uses the "Marco Geoestaditico, febrero 2018", and it can be accessed using this link. (NOTE: If you click on this link, you will start downloading the Mexican shapefiles (~3Gb) right away.)

Daymet

Daily climate measures for precipitation, minimum temperature, and maximum temperature (from 1980 to 2017) are taken from NASA Daymet and you can access the data here). To download the raster files for North America, you need to create an account first.

Mexican Migration Project

You need to get access to the restricted data from the Mexican Migration Project, which includes geocodes for the communities available in the MMP data.