We run the analysis using make version 4.4.1

make data-clean will generate all of the intermediate files we need for the analysis. On my computer it just took 4 minutes to run all targets. This does not run two categories of targets: downloading data and the water ML pipeline. Those both take a while to run, and thus I recommend you just get their data outputs from dropbox for now, along with the data we downloaded manually---they are grouped together in the data/download-manual, data/download-script, and data/watershed directories.

I've set it up so that make data-raw will create all of the files in the data/raw directory, make data-clean creates all of the files in the data/clean directory, make water creates all of the files in the data/watershed directory, and make data-download creates all of the files in the data/download-script directory.

For analysis:

• make desc-figs will create the descriptive time series plots and maps
• make cnty-results runs the county level analysis (event studies, DiD, TSLS for many outcomes and different treatment vars) and creates the event study figures. It also does the Ag district level analysis
• TODO: The main analysis is not yet in the makefile, we need to add it!

The following API keys are required, save them to .Renviron with with usethis::edit_r_environ()

• USDA QuickStats API saved in .Renviron as NASS_KEY
• Census API saved in .Renviron as CENSUS_KEY
• BEA API saved in .Renvoron as BEA_KEY

Ag district to county crosswalk from here.