Difference between revisions of "American Community Survey (ACS) Data"

From edegan.com
Jump to navigation Jump to search
Line 48: Line 48:
 
The API doesn't seem to require a key, but you can request one from https://api.census.gov/.
 
The API doesn't seem to require a key, but you can request one from https://api.census.gov/.
  
−
The data contains null values, as well as values that indicate error codes (-999999999,-888888888,-666666666,-555555555,-222222222, and others. See https://www.census.gov/data/developers/data-sets/acs-1year/notes-on-acs-estimate-and-annotation-values.html. Generally, it appears that all negative and null values should be striped. Every field is string quoted 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 https://www.census.gov/data/developers/data-sets/acs-1year/notes-on-acs-estimate-and-annotation-values.html. Generally, it appears that all negative values can be striped. Every field is string quoted including error codes, except for nulls.  
  
 
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 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 syntax one line at a time, and completely ignores the structure of the data (and so 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, which makes using a JSON handler likely worthwhile (otherwise, one would just strip the JSON tags and treat it as a badly encoded CSV, which it is).
+
The API returns improper JSON. Essentially, it just encodes a flat data structure into JSON 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.
  
 
To request a group ask for '''group(B06009)''' where you would normally request a field like '''B06011_001E'''. Programmers can rely on groups and fields being identifiable from the "_" and the string length (\d{5}\w{0,1}_\d{3}\w{1}).
 
To request a group ask for '''group(B06009)''' where you would normally request a field like '''B06011_001E'''. Programmers can rely on groups and fields being identifiable from the "_" and the string length (\d{5}\w{0,1}_\d{3}\w{1}).

Revision as of 19:16, 11 February 2021


Project
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 edegan.com. All Rights Reserved.


See also:

Notes

The ACS releases datasets in October of the following year. It has 1-year estimates or 5 year estimates. The five year estimates have more information.

The 1 yr estimates are only available for places with populations of 65,000 people or more, the 1yr supplemental estimates cover down to places with 20,000 people or more, and the 5 yr estimates cover a whopping 29,573 places [1], as compared with just 630 for the 1yr and 2,323 or the 1yr supplements.

One can download the PUMS data for 2017 here: https://www2.census.gov/programs-surveys/acs/data/pums/2017/1-Year/ However, PUMS is only available for places in the Public Use Microdata Areas, which have a population 100,000 or more.

And state-by-state, it looks possible to get the latest data profiles (2013-2017 5 yr estimates) but not place by place as data, just as an aggregate: https://www.census.gov/acs/www/data/data-tables-and-tools/data-profiles/

For just population, also see: http://census.ire.org/data/bulkdata.html. This seems to be based on the 2000 or 2010 census data only.

Using api.census.gov

The new approach is to use api.census.gov to request information. Each year has an API URL, which specifies the dataset and that takes get and for name-value pairs that specify variables and geographies, respectively. Generally, one can request geography, variable, and examples HTML pages on the same URL (rather than specifying the params). The ACS5 data goes back to 2009 and the ACS1 goes back to 2005 using the base URL: https://api.census.gov/data/yyyy/acs/acsx/, where yyyy is year and x is 1 or 5.

For, example, for 2019, the URLs are:

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

Other links:

Other datasets available using api.census.gov are listed here: https://www.census.gov/data/developers/data-sets.html.

Working with the API

The API doesn't seem to require a key, but you can request one from https://api.census.gov/.

The data contains null values, as well as values that indicate error codes (-999999999,-888888888,-666666666,-555555555,-222222222, and others. See https://www.census.gov/data/developers/data-sets/acs-1year/notes-on-acs-estimate-and-annotation-values.html. Generally, it appears that all negative values can be striped. Every field is string quoted including error codes, except for nulls.

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 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.

To request a group ask for group(B06009) where you would normally request a field like B06011_001E. Programmers can rely on groups and fields being identifiable from the "_" and the string length (\d{5}\w{0,1}_\d{3}\w{1}).

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: 

E:\projects\agglomeration\ACS\CleanFiles

The Load.sql script was run against Vcdb4

Python Script to pull certain columns from excel file:

  • E:\McNair\Projects\Agglomeration\ACS_Downloaded_Data\pullCertainColumns.py

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 https://factfinder.census.gov/faces/nav/jsf/pages/download_center.xhtml
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

S1501 EDUCATIONAL ATTAINMENT
S2301 EMPLOYMENT STATUS
B01003 TOTAL POPULATION
B02001 RACE
B07201 GEOGRAPHICAL MOBILITY
B08303 TRAVEL TIME TO WORK
B19013 MEDIAN HOUSEHOLD INCOME
B19053 SELF-EMPLOYMENT INCOME IN THE PAST 12 MONTHS FOR HOUSEHOLDS
B19083 GINI INDEX OF INCOME INEQUALITY
B25003 TENURE
B25105 MEDIAN MONTHLY HOUSING COSTS
B28011 INTERNET SUBSCRIPTIONS IN HOUSEHOLD
G001 GEOGRAPHIC IDENTIFIERS

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:

  • Geo.id 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
Retrieved from "http://www.edegan.com/mediawiki/index.php?title=American_Community_Survey_(ACS)_Data&oldid=48420"