A Python package for collecting American Community Survey (ACS) data and associated geometry from the Census API in a pandas dataframe.

• Installation
• Quickstart
• Geometry
  • Points
  • Polygons
    • Shapefile resolution
    • Shapefile caching
• Publishing to Socrata
  • Credentials
  • Example: Create a new dataset
  • Example: Replace rows in an existing dataset
  • Example: Create a new dataset from multiple queries
• Troubleshooting
  • Clearing the cache

autocensus requires Python 3.7 or higher. Install as follows:

pip install autocensus

To run autocensus, you must specify a Census API key via either the census_api_key keyword argument (as shown in the example below) or by setting the environment variable CENSUS_API_KEY.

from autocensus import Query

# Configure query
query = Query(
    estimate=1,
    years=[2017, 2018],
    variables=['DP03_0025E', 'S0103_C01_104E'],
    for_geo='county:033',
    in_geo=['state:53'],
    # Optional arg to add geometry: 'points', 'polygons', or None (default)
    geometry='points',
    # Fill in the following with your actual Census API key
    census_api_key='Your Census API key'
)

# Run query and collect output in dataframe
dataframe = query.run()

Output:

autocensus supports point- and polygon-based geometry data for many years and geographies by way of the Census Bureau's Gazetteer Files and Cartographic Boundary Files.

Here's how to add geometry to your data:

Point data from the Census Bureau's Gazetteer Files is generally available for years from 2012 on in the following geographies:

• Nation-level
  • urban area
  • zip code tabulation area
  • county
  • congressional district
  • metropolitan statistical area/micropolitan statistical area
  • american indian area/alaska native area/hawaiian home land
• State-level
  • county subdivision
  • tract
  • place
  • state legislative district (upper chamber)
  • state legislative district (lower chamber)

Example:

from autocensus import Query

query = Query(
    estimate=5,
    years=[2018],
    variables=['DP03_0025E'],
    for_geo=['county:033'],
    in_geo=['state:53'],
    geometry='points'
)
dataframe = query.run()

Polygon data from the Census Bureau's Cartographic Boundary Shapefiles is generally available for years from 2013 on in the following geographies:

• Nation-level
  • nation
  • region
  • division
  • state
  • urban area
  • zip code tabulation area
  • county
  • congressional district
  • metropolitan statistical area/micropolitan statistical area
  • combined statistical area
  • american indian area/alaska native area/hawaiian home land
  • new england city and town area
• State-level
  • alaska native regional corporation
  • block group
  • county subdivision
  • tract
  • place
  • public use microdata area
  • state legislative district (upper chamber)
  • state legislative district (lower chamber)

Example:

from autocensus import Query

query = Query(
    estimate=5,
    years=[2018],
    variables=['DP03_0025E'],
    for_geo=['county:033'],
    in_geo=['state:53'],
    geometry='polygons'
)
dataframe = query.run()

By default, autocensus will attempt to fetch almost all shapefiles at a resolution of 1 : 500,000 (500k). Some sources among the Cartographic Boundary Shapefiles are also available at the lower resolutions of 1 : 5,000,000 (5m) or 1 : 20,000,000 (20m). To attempt to download a shapefile at a specific resolution, pass a value to Query's optional resolution parameter:

from autocensus import Query

query = Query(
    estimate=5,
    years=[2018],
    variables=['DP03_0025E'],
    for_geo=['county:*'],
    in_geo=['state:53'],
    geometry='polygons',
    # Optional arg to set a specific resolution: '500k', '5m', or '20m'
    resolution='20m'
)

Setting a specific resolution is only supported for polygon-based geometry.

To improve performance across queries that include polygon-based geometry data, autocensus caches shapefiles on disk by default. The cache location varies by platform:

• Linux: /home/{username}/.cache/autocensus
• Mac: /Users/{username}/Library/Application Support/Caches/autocensus
• Windows: C:\Users\{username}\AppData\Local\socrata\autocensus

You can clear the cache by manually deleting the cache directory or by executing the autocensus.clear_cache function. See the section Troubleshooting: Clearing the cache for more details.

If you have publishing permissions on a Socrata domain, you can publish your query results directly to Socrata via the method Query.to_socrata. This method uses socrata-py to upload your dataframe with the appropriate field types, formatting, and metadata.

You must have a Socrata account with appropriate permissions on the domain to which you are publishing. By default, autocensus will look up your Socrata account credentials under the following pairs of common environment variables:

• SOCRATA_KEY_ID, SOCRATA_KEY_SECRET
• SOCRATA_USERNAME, SOCRATA_PASSWORD
• MY_SOCRATA_USERNAME, MY_SOCRATA_PASSWORD
• SODA_USERNAME, SODA_PASSWORD

Alternatively, you can supply credentials explicitly by way of the auth keyword argument:

auth = (os.environ['MY_SOCRATA_KEY'], os.environ['MY_SOCRATA_KEY_SECRET'])
query.to_socrata(
    'some-domain.data.socrata.com',
    auth=auth
)

# Run query and publish results as a new dataset on Socrata domain
query.to_socrata(
    'some-domain.data.socrata.com',
    name='Median Commute Time by Colorado County, 2013–2017',  # Optional
    description='1-year estimates from the American Community Survey'  # Optional
)

# Run query and publish results to an existing dataset on Socrata domain
query.to_socrata(
    'some-domain.data.socrata.com',
    dataset_id='xxxx-xxxx'
)

from autocensus import Query
from autocensus.socrata import to_socrata
import pandas as pd

# County-level query
county_query = Query(
    estimate=1,
    years=range(2013, 2018),
    variables=['DP03_0025E'],
    for_geo='county:*',
    in_geo='state:08'
)
county_dataframe = county_query.run()

# State-level query
state_query = Query(
    estimate=1,
    years=range(2013, 2018),
    variables=['DP03_0025E'],
    for_geo='state:08'
)
state_dataframe = state_query.run()

# Concatenate dataframes and upload to Socrata
combined_dataframe = pd.concat([
    county_dataframe,
    state_dataframe
])
to_socrata(
    'some-domain.data.socrata.com',
    dataframe=combined_dataframe,
    name='Median Commute Time for Colorado State and Counties, 2013–2017',  # Optional
    description='1-year estimates from the American Community Survey'  # Optional
)

Sometimes it is useful to clear the cache directory that autocensus uses to store downloaded shapefiles for future queries, especially if you're running into BadZipFile: File is not a zip file errors or other shapefile-related problems. Clear your cache like so:

import autocensus

autocensus.clear_cache()