Things You Need to Know
- There is no long form this time. Those of you who were around for the last census or two are familiar with the
fact that the decennial census had two different questionnaires: a short form with a relatively small number of questions which everybody had to answer, and a long form with a lot more questions which was sent to a sample (about one in six) of households. The long form was the one where they asked the questions regarding your income,
education, occupation, how and where you travel to work, if you have any disabilities, etc. This was sometimes referred to by veteran statisticians and journalists as "the good stuff". Since there is no long form in 2010 we cannot expect to get any data products containing any of this "good stuff". For those of you familiar with Census Bureau data product nomenclature and acronyms, this means that there will not be a Summary File 3 ("SF3") this time. (Nor an SF4, etc.)
- Why not? Because while the good stuff was very good, it became less good as the decade following each census went on. It got a little old. So, for example, by the time we got new 2000-census SF3 data in 2002 the SF3 data from 1990 was 12 years old. It was decided that what was needed was a new "continous measurement" approach which would involve doing a survey with a questionnaire similar to the decennial census long form that would be administered to a sample of the population througout the decade. This idea came to be implemented over the last decade or so as the American Community Survey (ACS).
To get congressional support for the ACS the Bureau had to promise that they would recover some of the ACS cost by not doing a long
form in the decennial census any more. (There were other issues involved in this decision, such as "respondent burden", but we are over-simplifying here for the more casual user.) What this means is that if you are looking for poverty rates for census tracts in Cincinnati (for example) then you have come to the wrong place. Poverty data are based on questions regarding people's income and these are (were) long-form questions. The data published based upon questions that only appeared on the long form which was only administered to a sample of the population were referred to as "sample data". So another way of saying what we are saying here is that there is no sample data to be found here in the 2010 decennial census section. For that sort of information, you'll need to go to the American Community Survey section.
- There is an important (but complex) link between the decennial and ACS data. In a certain sense the decennial censuses of past decades could be looked at as two separate surveys: the short-form-only survey and the long-form survey. But they were very closely tied and the sampling issues invovled in processing the long-form data were relatively simple since the short form data allowed the Bureau to know exactly not only on the households (and group quarters) that were sampled, but also how large the sampling universes were. Although there were differences in the basic counts for a census tract reported on SF1 (short form data) vs SF3 (long form, or sample data) the differences were rarely non-trivial. There was a single common sampling universe based on a single point in time (Apri 1 of the decennial census year) and shared rules regarding residency and other fine points of where (e.g. where to count college students and snow bunnies) and how much to count the data (weighting, based on sampling factors). Most of that straightforward linkage and consistency is gone now. The relationship of the decennial data and the ACS data is only vaguely understood by all but a handful of statisticians who (we hope and assume) really comprehend the Census Bureau's complex use of controls and weighting to produce ACS data estimates. An oversimplified summary: the ACS uses population "controls" to complete their estimation process. Those controls are based on the post-censal population estimates by age, race, sex and hispanic origin for the survey period, and those estimates are based to a considerable extent on the most recent census counts.
- Why use census data if the ACS contains all the "Good Stuff"? If you need data for very small geographic areas, such as census blocks or block groups, small cities or townships, etc. then the decennial data may be your only viable source. Block data are not available from the ACS, and block group data are very limited; not to mention that they are based on five year periods, usually come with large degrees of sampling error and (n the case of block groups) are not available via the MCDC's ACS Profile app.
So if you want to know how many children or senior citezens live in a neighborhood you are going to have to go with the decennial data (even when it starts aging over the decade). Decennial data (when available) are better because they are actual counts rather than sample estimates. While you can use the ACS to get updated population data for entities as small as block groups through the decade, these will be statistical estimates based on a varying and sometimes inadequate sample size.
Of course, if your grant application requires you to report the number and percentage of poor persons in your area then you have no choice but to go with data from the ACS since the information is no longer available based on the decennial census.
Note: To access any of the data referenced below (and some others not referenced here) using the MCDC's Uexplore/Dexter software go to
the Decennial Census 2010 section of the Uexplore directory page, where you will find links to the relevant data directories.
To learn more about the 2010 Census visit the
Census Bureau 2010 Census web site.