I would imagine I can find better ways of getting a census dataset indexed by GEOID, but I settled on using the 2015 ACS, which has a web API. This survey has thousands of fields available, all indexed by GEOID.

1. Make the directories ./data/ and ./data/json/. These will not be synced by the repo.
2. Go here and get an API key for the census API. Check your email and copy the key.
3. Go to ./build_census_table/ and run the command python query_acs.py <API key> population income household tenure marriage education, where <API key> is the key you received from the census bureau. Go get coffee while this runs (It should take a few minutes).
4. Run the command python build_table.py population income household tenure marriage education to build the table. This will not take as much time.
5. Boot up the jupyter notebook Rename Columns.ipynb and execute it. This renames the columns to something sensible.

Shape files for every census tract in the US can be found here. I added a script in ./data/shapefiles to scrape this page for you - all you should have to do is run python get_shapefiles.py. It is written for mac, but could easily be changed to work on linux. I plan to use the shapefiles for some fun visuals, but I havenn't planned out anything yet.

1. GEOID. See here
2. ACS. The American Community Survey. See here. We are using the 2015 5 year estimates to build our dataset.