Census 2010 - Less Local US Data & ACS Insufficient

Census 2010 is right around the corner, folks.

Have been hearing numerous faculty and students anxiously awaiting for fresh new Census data, as the 2000 data has gotten quite stale and the taste of marketing estimates is never as satisfying. However, the joyous party times we experienced during the early 2000's as summary files 3 and 4 were being released are not to be repeated this time around. As I will discuss below, this has huge implications for all users of demographic data and this in turn will have a huge affect on my role as GIS Librarian.

Census 2010: No Long Form

So, the US Census has gone and submitted the subjects for the 2010 Census to Congress. There will be only 6 subjects. The data for these subjects is planned to be released on the Census block level on April 1, 2011. Here are the 6 subjects:

1. gender
2. age
3. race
4. ethnicity
5. relationship (to the head of household)
   
6. whether you own or rent your home (tenure)

What?

How about the long form questions? How about income, education, citizenship, all those housing characteristics? Since 1960, sample data has been collected on the Census long form sent out to 1/6 households containing oodles of data. Data that I and many others have come to rely upon.

There will be no long form questionnaire in 2010. To cut costs, the US Census Bureau is now relying on the American Community Survey (ACS) instead of the decennial census to provide all questions previously on the long form.

This is not good news, especially in the short term. I have a big problem with the ACS replacing the long form data previously released each decennial census since 1960. The implications scare me.

Let's Take a Closer Look at the ACS

The American Community Survey last year began providing access to annual estimates of various demographics for all cities (places), counties, congressional districts, school districts, and states in the US.

Now, the upside to the ACS as compared to long form decennial census data is the timeliness. No longer will we need to wait until next decade to find the median household income for our fair city of Arlington. The ACS will release this data annually.

The downsides, however, are quite steep. First, the ACS is currently releasing data down to the county and city level. It is extremely rare that our students here use census data on any level other than block, block group, tract, or zip code tabulation areas. However, according to Chris Williamson, Ph.D., Senior Planner for the City of Oxnard, California, on the latest Planetizen Podcast (which is efficiently informative), as soon as the ACS has accumulated enough data they will begin releasing data on the tract level. In am uncertain when. Regardless of when, we have all been enjoying long form data down to the block group level.

Second, the margin of error for ACS data can be quite high. When using ACS data, it is imperative to consider the margin for error that is provided for each attribute for each geography. Now, Dr. Williamson (Planetizen Podcast) stated that the accuracy of ACS data will improve over time, but for the short-term this is what we have.

ACS Data can be accessed via Factfinder, FTP download, and as data briefs (URL fixed)

Two Ways This Will Impact GIS At My Campus

1. Possibly stifle the use of GIS
   
1. Over the last two years, GIS activity on campus has blossomed as more and more faculty and students are embracing geospatial technologies to supplement their research. For many new social science, social work, health care, urban planning, and business users, the ease of accessing census data and joining it to TIGER shapefiles is a huge draw. The ease substantially decreases if the new users must consider the margin of error before using ACS data.
2. Losing the ability to allow users to analyze official census data on the block group level (beyond the basic six subjects). As I discussed above, ACS is currently available to the city (place) level, and there are plans to release data on the tract level.
   
2. Increased reliance on unofficial and less accurate estimates
1. Faculty and students will not be willing to give up their block group level data. We (the library) will have to provide it to them. I assume (hope) companies such as Applied Geographic Solutions will continue to create current and 5 year estimates, and of course we will continue to collect this unofficial data. However, if the census is publishing data only down to the city or tract level, and if the margin for error is higher than ever before, the error of these unofficial estimates will increase. At the same time our users' reliance on these unofficial datasets will increase as ACS is insufficient to satisfy their needs. Perhaps the timeliness of the ACS will help to counteract this.

---

Now, I understand the Census Bureau (which means us, the taxpayers) is saving money by shifting from the decennial long form to the American Community Survey, but is it really worth it. Perhaps if I knew the amount of the savings it would make more sense to me. There are a whole lot of bright folks working at the census, and I reckon if they decided this is the best thing to do, then perhaps it is.

Free Historical Census Data: 1970, 1980, 1990

Received a request over the weekend for 1990 Census data on the block level beyond what is currently available through our local holdings. We are purchasing the CensusCD 1990 Blocks from Geolytics, but in the meanwhile this forced me to take some time exploring free historical Census data availability and so here are the results of what I discovered.

U.S. Census Bureau

• Data Files: Census 1980
  
• Data Files: Census 1990
• These downloadable data files are fantastic in that users can freely download the entire 1980 and 1990 dataset, including all the summary tape files (STF). Downside is they are not easy to use. Choice is to install the Extract software or access the DBF files yourself. In the past I imported the DBF file containing the STF 1B for Texas into an Access database and made that database available to our library users, who would then run queries against the data. (Access, of course, will be much improved when we acquire CensusCD 1990 Blocks.
• American Factfinder
• Everyone loves this fantastic resource, but it only offers data from 1990 and 2000. Their 1990 holdings do not contain STF 1B, and so do not go down to the block level.

University of Wisconsin-Madison Center for Demography and Ecology

• 1970 Census Summary Tape Files
• Files for all available states and the District of Columbia for the 1970 4th Count A and C population and housing files may be downloaded.
• Current Population Surveys (various, from 1963 - 1990)
• 1980 Public Use Microdata Sample (PUMS)
  

UC Berkeley Social Science and Government Data Library: Files to FTP

• Been having trouble accessing their public archive, which is too bad as it has been a fantastic resource. Seems as if their FTP server ftp://sunsite2.berkeley.edu/pub dis not permitting public anonymous logon. Their FTP server ftp://sunsite.berkeley.edu/pub is permitting anonymous logon, but does not contain their data. Hopefully their data will be made available to the public again.

Interuniversity Consortium for Political and Social Research (ICPSR)

• This is not a free resource, but it seems silly comprising this list without mentioning the ICPSR. Many (if not most) academic universities should be a member of ICPSR and have access to its data holdings.
• Here are their Decennial Census holdings.