Questions/concerns relating to the Census Bureau's count resolution, please see the Federal Register Notice, Monday, January 22, 2001, Vol. 66, No. 14.
The files distributed by the Census Bureau provide a set of four very detailed tables (containing more than 280 data cells per area) describing the population of a wide variety of geographic areas. (See Chapter 6 of the Tech Doc for details.) These files are intended to be used in redistricting applications which typically involve the use of very sophisticated software packages. These files can tell you such things as the number of people who checked any of 63 combinations of racial questions on their census form within geographic areas as small (and incomprehensible without a good map) as block groups and census blocks.
Somewhat hidden within the complexity of these complete tables are relatively simple population counts. While we have (or will have shortly — depending on what day it is) data files available at the MCDC with all 280+ data cells for all geographic areas in Missouri and some other states, what we are emphasising is a standard extract version of the file that we have created from the much larger files. These extracts have data on basic population counts, broken down by race, hispanic origin and age (all counts are available for both the total population and the population age 18 or over).
PL94 is also complex in terms of the specialized census geography that us used to tabulate the data. For complete details see the Tech Doc, Chapter 4 - Summary Level Sequence Chart. If you are a normal human being, not directly involved in a redistring program, there is a very good chance that 90% of the geographic areas summarized on PL94 would be of no interest or use to you. With this in mind, we have attempted to create data summaries that emphasize the most commonly used geography (states, counties, cities, mcds, census tracts and block groups). Our geographies are all the complete versions as opposed to the split versions which occupy most of the space on the PL94 files. Most people wanting block group data want it for the entire block group, not just the portion within a VTD (whatever that is), within a place and within a county subdivision.
For Missouri (at least) we have also been very interested in getting comparable data (where possible, or nearly comparable in some cases such as racial data) from the 1990 PL94 files. Our primary means of delivering these extracts will be in the form of a series of electronic reports accessible from this web site. Many of these reports have been (or soon will be) published in hard copy format and have been disseminated around the state to our affiliate and core agencies as well as to government document libraries.
If your interest requires access to the complete 63 race combinations and/or geographic detail involving the portion of a block group within a voting district, mcd and place (for example), then you are probably better off going to the Bureau's web site. You can either look for data on the American Fact Finder page or you might want to purchase a cd-rom for your state (only $50) — comes with excellent software to access and display very detailed (and even not so detailed) tables.
The thorniest issue that data providers and users have to deal with related to the new PL94 data (and this is going to be with us for all 2000 data products) is how to deal with the change in race categories. In earlier decades the census questionnaire asked persons to check a box indicating their race. They were instructed to check only one such box. Everybody was either white or black or Japanese or American Indian, etc. But an OMB directive from 1997 has mandated that all federal surveys should now allow respondents to be able to specify multiple races. You can check as many of the race boxes as apply. For the sake of tabulating the data for PL94, there were six basic race categories (up from five basic categories in 1990; the "Asian and Pacific Islander" category used in 1990 has been split into two categories for 2000: "Hawaiian and Other Pacific Islander" and "Asian".) You can view the results of this new way of assigning race categories when you look at the size and complexity of the four tables on PL94.
Although the additional race detail gathered by allowing the multiple responses is useful for someone studying racial mixing patterns, there is a real problem for most applications of the data. If it were a matter of a single table with counts of the 63 combinations it would be fine. But these 63 categories are not just a single table in the census tabulations. Many census tables (including all of them on PL94) use race as a table dimension. So we have tables of "Race by Age" and "Race by Sex" and "Median Household Income by Race of Householder", etc. Do we really want all of these tables to give us the full 63-category detail? Is more data better? Not in most cases. The data need to be further collapsed to make them useful for most applications.
One way that has been suggested (initially by the Census Bureau) to deal with the new complexity was to summarize race data by looking at two basic counts associated with a given basic racial group. If you want to look at data for the Asian racial category (for example) you would get data for all persons who indicated that they were Asian and no other race, i.e. "Asian alone". The companion category would be all persons who checked the "Asian" box on their form, i.e. "Asian alone or in combination with other races". These two counts represent an interval in which the value of "Asian persons" falls. The reason we have an interval instead of a simple count as before, is because now being "Asian" is not a simple yes/no question. You can be completely Asian ("Asian alone" — the lower limit of the interval) or you can be at least partly Asian ("Asian alone or in combination..." — the upper limit of the interval.) Of course, you can also be in the partly Asian category; these are the people who make up the interval.
In producing products based on the PL94 data, the MCDC has decided to focus primarily on the use of these racial intervals for report race-stratified data. It turns out that this was not exactly a trivial thing to do. If you study the 288 cells of the four tables on the PL94 files as distributed by the Census Bureau, you will not find any explicit counts of such things. To determine the number of persons who are "Asian alone or in combination.." you have to identify and sum the 32 cells in Table PL1 that indicate persons who are all or partly Asian. Trying to then determine how many of this total is of voting age and/or of Hispanic origin requires even more careful study and aggregation of data in the other PL tables. The process is tedious (even by our standards) and error prone. But we felt it was worth the effort, so we created our own transformed extract files where we created exlicit counts that were based on the "racial interval" approach.
We are creating a series of standard reports based on the data reported in the PL94 files. Each of these reports contains similar data in terms of the demographic characteristic reported. They vary only in terms of the geography involved -- what geographic entities are summarized and what geographic universe. So, for example, there is one such report for counties (and the state totals) in Missouri, and another for the same geographic entities in Illinois, and another for Kansas, etc. There is another report that shows Missouri places with names starting with "A" thru "D", etc. The reports also have a time dimension to them. They not only report results from the 2000 census but attempt to report comparable, or nearly comparable, data from the 1990 census. The reports can vary in terms of whether or not the 1990 data is reported. For small area geography (census tract and block group, for example) you need to convert the 1990 data from the census geographic units used for reporting the data in 1990 to the corresponding 2000 units. (Census tracts in general may change across decades.) As a result, we may not have had the resources to convert the 1990 data for other states (other than Missouri) to the 2000 equivalents. But we were able to do so for Missouri, so all reports for Missouri should include the 1990 data.
You will notice that in these reports there are racially-based "Population Group" counts which for 2000 are reported as a pair of figures separated by a "-". These numbers are the "racial intervals" described above. The corresponding data for 1990 has simple counts for the corresponding cells since racial categories were simpler then. We purposely opted not to explicitly attempt to report any value labeled as a change or percent change in any race-based category. People are going to want to know what the change was in the African American population of Missouri between 1990 and 2000 was. But we have no simple one-number answer. But we can provide any interval. For specific kinds of applications involving legal definitions of race it may be possible to cite a more specific answer to the question of how many persons are African American. But there is no single universal one-number answer.
The MCDC has attempted to do what we can to make it easy for data users to get trends in these PL94 data itesm over time. Specifically, we have been reworking data from the 1990 PL94-171 files. Data from 1990, reworked to be compatible with 2000 (and for use in our Basic Trend Reports) has been created within the MCDC data archive. These data sets can be accessed using our uexplore/xtract web applications. The directory containing the relevant 1990 data is /data/census_1990/pl9490/pl9490tx. The URL for invoking the uexplore application pointing to this directory is /cgi-bin/uexplore?/data/census_1990/pl9490/pl9490tx@secure. Note especially the data sets that have names with "00" in them. These are summaries of 1990 data by 2000 census geography. For example, the data set motrs00 has 1990 data tabulated to 2000 census tracts. This, of course, is what we'll be using to create tract-level Basic Trend Reports. But sophisticated data users will want to simply capture these data for processing within their own software packages.
Getting data for census blocks, block groups and tracts is only useful if you know where these geographic areas are. The key to this information lies in electronic maps prepared by the Bureau showing detailed street networks along with the boundaries and labels of the census geography. The Bureau will be posting this maps to the web in a series of pdf files. You can access these from the web. Hard copies of the maps for Missouri will be available from the Geographic Resources Center at the University of Missouri. We also understand that Kinkos and others can make hard copies from the pdf files. The maps are designed to print on large "E size" paper (36" x 34").