This code is part of a project (for me) to learn Python through the NYU-Stern-Econ databoot camp course (http://databootcamp.nyuecon.com/). It combines US 2016, 2012 Election data at the county level with Census data on education and earnings, and then the Autor, Dorn, and Hanson (2013) US Trade Exposure from China data.
This code does the following: It imports the 2016 Election Data, 2012 Election, imports a Country FIPS code to Commuting Zone mapping and then merges it with Census data on education and earnings, the ADH data which measures the exposure of a commuting zone to Chinese trade.
It then correlates the change in the republican vote with education, earnings, population size, and the trade exposure measure.
Preliminary results: Education and Population Size explain 60% of the change in the republican share. Lower education and rural areas are more likely to change towards Trump relative to Romney.
This relies upon four public data sets that are in the site, but can be found here:
-
2016 US election data : https://data.world/aaronhoffman/us-general-election-2016
-
2012 US election data: https://data.world/aaronhoffman/us-general-election-2012
-
Census, American Community Survey: https://www.census.gov/programs-surveys/acs/
-
County FIPS to Communting Zones: http://www.ers.usda.gov/data-products/commuting-zones-and-labor-market-areas/
-
Autor, Dorn, and Hanson data: http://www.ddorn.net/data.htm