Census 2000: An overview
By Matthew Waite
Times Staff Writer
In order to overcome changes in the way the Census was taken in 1990 and 2000, the St. Petersburg Times used mapping software and statistical analysis to produce the graphics and maps in the printed newspaper and online.
Each Census changes. The wording for the questions is a multiple-year process with each question reviewed by Congress. And, less noticeably, the areas the Census creates summaries for changes. In 1980, the Census Bureau hadn't created Census tracts -- areas of about 4,000 people -- for the entire country. In 1990, the bureau hadn't drawn lines for census blocks, much like city blocks.
Each change had an effect on how the data was analyzed.
The largest obstacle was that in 2000, people could mark more than one race. That meant that instead of five race categories, there were 63.
To make comparisons to 1990, the Times used a method developed by Arizona State University Professor Stephen Doig, which estimates what 2000 results would look like with 1990 categories.
The method divides multi-race responses into their 1990 categories. For example, a census tract with 60 people who recorded themselves as black, Asian and American Indian would be divided up into 1990 categories as 20 blacks, 20 Asians and 20 American Indians.
One exception to that is mixed-race responses that includes white. A tract with 40 people marking white, black and Asian would be counted under the Times method as 20 black and 20 Asians. The reason is because for any official counting of people where race is important -- such as voting rights act or affirmative action --mixed race responses wouldn't be counted as white.
The analysis only estimates what 2000 race responses would look like in 1990. The Census Bureau has said they intend to study how mixed race respondents answered the race question and create their own method. However, that method isn't expected for several years.
What are tracts and how are they used?
To do local analysis, the Times used census tracts, areas drawn by the bureau that represent about 4,000 people. Between 1990 and 2000, however, the bureau redrew some lines and divided some tracts into smaller sizes.
In order to compare local numbers from 1990 to 2000, the Times used a process where 1990 block counts were added up into the 2000 census tracts.
Where a block was divided by a tract line, the block was divided proportionally to how the block was divided. A block with 10 people divided in half would assign five people to one tract, five to the other.
This method estimates what 2000 tracts would have held in 1990.
What are Places?
The Census Bureau also provides data for what they call Census Designated Places. Those can be cities, like St. Petersburg or Brooksville, and they can be unincorporated places like Westchase or Holiday.
Places can change from census to census. Some new places are added, old places are dropped, and lines are expanded.
For the Times analysis, 2000 places were used, leaving out places that were counted in 1990 but not in 2000. Places that were counted in the 2000 Census for the first time will include zeroes for 1990 numbers, and therefore would not include any percent change calculations. That does not mean, however, there were no people living within those boundaries 10 years ago. Areas that are not designated as places are left unbounded and show up in white on the map. For information on those area, see the tract maps.