Make workflow for downloading Census geodata and joining it to survey data
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.

Get tiger

A make-based tool for downloading Census Tiger Line Shapefiles and automatically joining them to data downloaded from the Census API. If you want to make maps with US Census data, but find downloading it a pain, this is for you.

Get-tiger uses make, a tried-and-true tool for processing series of files, to quickly download Census geodata and survey data, then join them. Then you have Shapefiles (or GeoJSON) ready to use in your favorite GIS.


  • Make (tested with GNU make 3.81, other versions should work fine)
  • ogr2ogr (GDAL) (v1.10+)

GDAL is an open-source geospatial library that includes ogr2ogr, a commmand-line tool for modifying GIS data. We'll be using it to join CSVs to Shapefiles.


  • install make with: xcode-select --install.
  • For GDAL, install Homebrew and run: brew install gdal.


Linux (CentOS):

  • sudo apt-get install build-essential g++ libgdal1-dev gdal-bin


  • Download or clone the repo and put the contents in the folder you would like to fill with GIS data.
  • Get a Census API key (yes, it's pretty bare-bones).
  • Put that key in key_example.ini, and rename it key.ini.


Running make will produce a list of Census geographies available for download:

Available data sets:
Download with make DATASET
NATION - United States
UNSD - Unified school districts
ZCTA5 - Zip code tabulation areas

To download one or more, run with name of the dataset, like so:

# Download the national and state files and data
# Download county data and each state's tract data

Make will run the commands to download the shapefiles and data from the Census, then join them. You'll see the commands run on your screen, sit back and enjoy the show. The files will end up in a directory called 2014/. At the end, you'll get a list of the files created, e.g.:

2014/NATION/cb_2014_us_nation_5m.shp 2014/STATE/tl_2014_us_state.shp

Some commands will download many files. For instance, this will download files for the fifty states, DC and Puerto Rico:

make PLACE

To download only some states and territories, use the STATE_FIPS variable:

# Only New York

# Only DC, Maryland and Virginia
make PLACE STATE_FIPS="11 24 51"

You may find a list of state fips codes handy.

Which maps

The Census publishes two sets of map data: Cartographic Boundary files and TIGER/Line. The main difference is that cartographic boundaries files are clipped to the coastline. These are the default for get-tiger. To always fetch TIGER/Line files, set CARTOGRAPHIC=false:


What data

A current weakness is that data is downloaded with no data dictionary, and cryptic field names. I've included a data dictionary (data.json) for the default fields.

To download different data, see the Census API documentation for a complete list. You must select each field separately. Due to the way the Census API is set up, one cannot just download an entire table.

Make your selection, then run it, e.g.:

make STATE DATA_FIELDS="B24124_406E B24124_407E"

You could also add these fields to key.ini:

DATA_FIELDS = B24124_406E B24124_407E

This will override the defaults in the Makefile.

Sometimes you want to make summaries based on the raw data fields. For instance, you might want to divide population by land area to get population density. The variable OUTPUT_FIELDS can be used to add this kind of summary field.

This example adds a field called TransitCommutePct, which is produced by dividing B08101_025E (estimated number of workers who commuted by transit) by the B08101_001E (estimated number of workers)

OUTPUT_FIELDS = B08101_025 / B08101_001 AS TransitCommutePct,

Note that in the expression, the E is dropped from the variable names. This is due to a limitation of the Shape format - only the first 10 letters of field names are used. Also, the expression must end in a comma.

Another example:

OUTPUT_FIELDS = B01003_001/(ALAND/1000000.) AS PopDensityKm,

The field B01003_001 is population. The census calls it B01003_001E, but has been shortened to 10 characters. We divide it by ALAND/1000000., because the ALAND field is given in meters. The AS PopDensityKm gives the name of the new field. Finally, it must end with a comma (,).


By default, the Makefile downloads 2014 data, the most recent year for which ACS is available. For older years (or newer years, if it's the future), use the YEAR variable:

make STATE YEAR=2013
make STATE YEAR=2015 

Data series

The most recent 5-year data available on the API is from 2014. Data from 2015 is available, but only from the 1-year dataset. To fetch another data set, use the SERIES variable. The default is acs5, use acs1 for 1-year data.

make TRACT SERIES=acs1 YEAR=2015

Secret bonus tasks for merging data

A relatively common task is to download a national set of geographies of a certain type. Run this to download a national dataset of block groups:

make 2014/BG.shp

You can add in options for different DATA_FIELDS as described above. To run this task for a different year, you'll need to change the year twice (make 2013/BG.shp YEAR=2013).

Get-tiger includes shortcut tasks like this for the following geographies. They all follow the same pattern (2014/<NAME>.shp):

  • American Indian / Alaska Native Areas / Hawaiian Home Lands (AIANNH, AITSN, ANRC)
  • block groups and tribal block groups (BG, TBG)
  • blocks (TABBLOCK)
  • census tracts and tribal census tracts (TRACT, TTRACT)
  • congressional districts (CD)
  • consolidated cities (CONCITY)
  • counties (COUNTY)
  • counties within urban areas (COUNTY_WITHIN_UA)
  • county subdivisions (COUSUB)
  • metropolitan areas (CBSA, CSA, METDIV)
  • military bases (MIL)
  • New England stuff (CNECTA, NECTA, NECTADIV)
  • places (PLACE)
  • primary/secondary roads (PRISECROADS)
  • public use microdata areas (PUMA)
  • Puerto-Rico-specific subdivisions (ESTATE, SUBBARRIO)
  • railroads (RAILS)
  • roads, primary roads and primary plus secondary roads (ROADS, PRIMARYROADS, PRISECROADS)
  • school districts (UNSD, ELSD and SCSD)
  • states (STATE)
  • state legislative districts (SLDL, SLDU)
  • urbanized areas (UAC)
  • zip code tabulations areas (ZCTA5)
  • high-level subdivisions (NATION, DIVISION, REGION)


By default, this thing spits out Shapefiles. To get GeoJSON, set format to json:

make TRACT format=json
make TRACT format=shp # default
make 2014/COUSUB.json format=json


Get-tiger will integrate well with just about any workflow. For instance, here's a basic make recipe to automatically download the repository and then get data county-level data:

# Census API key
export KEY=12345 
# chosen data fields
export DATA_FIELDS="B24124_406E B24124_407E"

# If you do this, you should probably fork the repository and clone your fork
    git clone --single-branch census-data
    touch census-data/key.ini # You'll need to create the key.ini file to prevent an error
    $(MAKE) -C census-data COUNTY

Your data will be available in census-data/2014/COUNTY. These steps can be readily performed by the scripting language of your choice.

Interesting tidbits

  • The Census API appends extra geography fields at the end of a request. For example, 'state', 'county', and 'tract' for a tract file. As part of the processing, these are converted to numeric values, which reduces their usefulness. Use the GEOID field for joining.
  • The AWATER (water area) and ALAND (land area) fields are given in square meters. The Shapefile format has trouble with values more than nine digits long, so these will trigger warnings in ogrogr. The Makefile adds LANDKM and WATERKM fields (the same data in square kilometers) to get around this issue. Also, get-tiger squelches the warning messages on these operations.
  • Where available, get-tiger will download the cartographic boundary files, rather than Tiger/Line files. The cartographic files are clipped to the shoreline, Tiger/Line files are not. If you would prefer the Tiger/Line files, open an issue and I'll add a way to download them.
  • Run tasks with the --jobs option (e.g. make --jobs 3) to take advantage of a fast connection and/or computer.
  • Downloading data for blockgroups, area water, linear water, or roads requires downloading data county-by-county. This means get-tiger needs a list of all the counties in the US. A list of 2013 & 2014 counties is included. If you're downloading blockgroups for other years, run make counties/2020.ini YEAR=2020 before running make BG YEAR=2020 or make AREAWATER YEAR=2020.


Copyright 2016 Neil Freeman. Available under the GNU General Public License.