American Community Survey (ACS) Data

Jump to navigation Jump to search

American Community Survey (ACS) Data
Project logo 02.png
Project Information
Has title American Community Survey (ACS) Data
Has owner Jeemin Sim
Has start date
Has deadline date
Has project status
Has sponsor McNair Center
Has project output Data
Copyright © 2019 All Rights Reserved.

See also:


The ACS releases datasets in September through December of the following year. It has the following datasets:

  • ACS5: 5-year estimates for a broad list of variables, from 2009-2020, covering 29,573 places.
  • ACS1: 1-year estimates from 2005-2020 for select variables for 630 places with populations of 65,000 people or more
  • ACS1s: 1-year supplements covering 2,323 places with 20,000 people or more
  • Some historical datasets (like ACS3), which are not currently supported

ACS provides information on a broad range of geographies, including states, places, and zctas.

Other data sources

Note: PUMS is only available for places in the Public Use Microdata Areas, which have a population 100,000 or more.


The new approach is to use to request information. This applies to ACS and a broad range of other Census datasets.

Each year has an API URL, which specifies the dataset and takes get and for name-value pairs that specify variables and geographies, respectively. The base URL for ACS is:, where yyyy is year and x is 1, 1s, 3 or 5. An example request is:

Pages listing available geographies, variables, and some examples are on the same base URL. For, example:

A script called automates the retrieval and compiling of results from It has the year, geographic unit, and dataset in a (user-editable) parameters dictionary and is called with variables as command line parameters. Note that groups of variables are called in the same way as individual variables.

The script is available in:


Example commands are:

python3 B19013_001E B06009 B02001
python3 B06011_001E 

The script produces tab-delimited text files (to change to csv, edit line 63) with timestamped filenames.

Note: In the script, I rely on the response.json() method of request to handle the "JSON" produced by the API. I never use json.loads or similar. The Census doesn't adhere to any reasonable standards for encoding URLs or data!

Examples and Documentation

Example API calls for median income, educational attainment, and race from the 5yr ACS, using the 2019 dataset:

Other links:

Other datasets available using are listed here:

Working with the API

Getting a Key

The API doesn't seem to require a key, but you can request one from

Quoting, Nulls, and Error Codes

Every field is string quoted (even if it's a number) including error codes, except for nulls. The data contains null values, as well as values that indicate error codes (-999999999,-888888888,-666666666,-555555555,-222222222, and others. See Generally, it appears that all negative values can be striped.

Improper Encoding

The API uses improper URL encoding (or more accurately, doesn't use URL encoding), so most methods for auto-populating URL parameters in scripting languages will not function correctly. Build URLs manually.

The API returns improper JSON. Essentially, it just encodes a flat data structure into JSON-like syntax one line at a time, and completely ignores the structure of the data (and so much of the point of encoding it in JSON!). However, the first (zeroth) column is the requested geography name, so multiple pulls can be joined using that. And the first (zeroth) row is a valid set of column headers. Also, the nulls are valid JSON encodings. This, and the strong quoting, makes using a JSON handler likely worthwhile. An alternative is to strip the JSON tags and treat the data as a badly encoded CSV, which it is.

Field and Group Names

To request a group ask for group(B06009) where you would normally request a field like B06011_001E. It looks like programmers can rely on groups and fields for estimates being identifiable from the "_" and the string length (\w{1}\d{5}\w{0,1}_\d{3}\w{1}), where the first letter is B or C and the last letter is E. Attributes for estimates can have two one last letter (like M for Margin) or two last letters(like AE - Annotation of Estimate and AM - Annotation of Margin). Note that requests of individual estimates don't provide the attributes but requests of groups do.

Old Approach using AFF

The approach below, using American FactFinder (AFF), was the old standard. Unfortunately, AFF has now been phased out. See the Fact Finder Download Center.

Old Files

All of the corrected files and the Load.sql script are in:


The Load.sql script was run against Vcdb4

Python Script to pull certain columns from excel file:

  • E:\McNair\Projects\Agglomeration\ACS_Downloaded_Data\

SQL Commands to create tables and load data:

  • E:\McNair\Projects\Agglomeration\ACS_Downloaded_Data\DataLoading_SQL_Commands

Excel files downloads from ACS website:

  • E:\McNair\Projects\Agglomeration\ACS_Downloaded_Data\csvFiles

Steps to Obtain Data

1) Go to
2) Select 'I know the dataset or table(s) that I want to download.'
3) Press Next
4) For 'Select a program:' choose
       'American Community Survey'
5) For 'Select a dataset and click Add to Your Selections:' choose
       '<YEAR OF INTEREST> ACS 1-year estimates'
6) Press 'Add To Your Selections'
7) Press Next
8) For 'Select a geographic type:' choose
       'Place - 160'
9) For Select a state:
       Don't choose a state, as we wish to download all.
10) For 'Select one or more geographic areas...' choose
       'All Places within United States and Puerto Rico'
11) Press Next

For the new pull, I used 2017 5yr Estimates.

Tables Pulled

B02001 RACE

At the end, I chose to have full descriptions but not geographic components.

Notes on Data

  • Educational Attainment are based on population of age 25 years or higher (except for years 2010, 2011, and 2012 - that's why there are fewer entries in those tables)

Loading to a Dbase

The zip file from AFF was downloaded and extracted to E:\projects\agglomeration\ACS\AFFFiles

There are 13 files entitled ACS_17_5YR_table_with_ann.csv, where table is S1501, etc. The files have two header lines and are CSV, with quoted strings of the format "Abanda CDP, Alabama" for the display label. There are 5 columns in the first file:

  • e.g., 1600000US0100100
  • GEO.id2 e.g., 0100100
  • GEO.display-label e.g., "Abanda CDP, Alabama"
  • HD01_VD01 (Estimate; Total) e.g., 174
  • HD02_VD01 (Margin of Error; Total) e.g., 158

Geo.id2 appears to match our Geoids. Note the leading zero, so these fields should be converted varchars(7) with front padding.

Checking Geo.id2 of 0103076:

SELECT * from Tigerplaces WHERE geoid='0103076';
statecode	statename	gid	statefp	placefp	placens	geoid	name	namelsad	lsad	classfp	pcicbsa	pcinecta	mtfcc	funcstat	aland	awater	intptlat	intptlon	geom
AL	Alabama	380	01	03076	02403132	0103076	Auburn	Auburn city	25	C1	Y	N	G4110	A	152375113	2646161	+32.6077220	-085.4895446	
1600000US0103076,0103076,"Auburn city, Alabama",61462,65

All files are renamed table.txt and have their second header line removed. The first three columns are common to all files, but then files can have multiple estimate and margin columns. Redundant ones are deleted. The state is also split out. The following are kept:

  • B01003 - Population - Total, margin
  • B02001 - Race - Total, WhiteAlone, BlackAlone, IndianAlone, AsianAlone, IslanderAlone, OtherAlone, TwoPlus
  • B07201 - Moving - Total, Same1yr, DiffinUS1yr, DiffinMSA1yr, Abroad1yr
  • B08303 - Driving - Total, Lt5, Btw5And9, Btw10And14, Btw..., Btw60And89, Gt90
  • B19013 - HHIncome - MedHHInc2017
  • B19053 - SelfEmployment - Total, NumWSelfEmpIncome, NumWOSelfEmpInc
  • B19083 - GiniIndex - Gini, margin
  • B25003 - Accomodation - Total, Owner, Renter
  • B25105 - Housing - monthlyhousingcost
  • B28011 - Internet - Total, InternetSub, DialUp, Broadband, Satellite, Other, NoSub, NoInternet
  • S1501 - Education - Total25plus, Male25Plus, Female25Plus, Assoc25Plus, Bach25Plus, Grad25Plus, HSOrHigher, BachOrHigher, PCPovertyDueToEdu, MedEarnings
  • S2301 - Labor - Total, LFPR, EmpPopRatio, UnEmpRate