Notebooks will power effort to verify census
Notebooks will power effort to verify census
By Patricia Daukantas
After most of the 2000 Census is complete, a separate team of workers equipped with notebook computers will hit selected streets to verify the decennial tally.
Census crew leader Kathy Henderson interviews a resident for the Census 2000 Accuracy and Coverage Evaluation Project.
The Census Bureau's review program, called the Accuracy and Coverage Evaluation, will re-count roughly 0.3 percent of U.S. housing units to give the bureau an idea of how much it undercounted the population. Past evaluations of the decennial census have shown chronic undercounting that varies by ethnic group, said Rajendra P. Singh, assistant chief of the bureau's Decennial Statistical Studies Division. The 1990 Census made a slight improvement on the decades-long trend of undercounting, Singh said.
For the verification program, the bureau divvies up the nation into blocks, or small geographical areas defined by visible boundaries. An ideal block will have about 30 housing units based on the Master Address File that the bureau compiled for Census 2000.
Statisticians will randomly select enough blocks throughout the 50 states and the District of Columbia to total about 300,000 housing units, plus another 15,000 in Puerto Rico. The verification project also will sample 10,000 to 11,000 housing units on American Indian reservations. Census found it had a large undercount on reservations in 1990, Singh said.
The evaluation will not include group quarters such as dormitories, prisons or military barracks, Singh said. It will also exclude remote areas of Alaska, which received special attention at the start of the Census 2000 effort.
After selecting the blocks, the bureau will make an independent address listing instead of relying on the Master Address File information.
'The intent is to keep this independent of the census,' said Jane Polzer, external liaison in the bureau's Technologies Management Office.
Late this spring, when the decennial census is almost completed, verification officials will hire temporary workers to visit each housing unit in the chosen blocks.
During their visits, the temps will conduct computer-assisted personal interviews, or CAPIs, typing residents' responses into the questionnaire program loaded on their Hewlett-Packard OmniBook 900 notebook PCs running MS-DOS 6.22.
The bureau stuck with MS-DOS because its data collection software, the Computer-Assisted Survey Execution System (CASES) from the University of California at Berkeley, was not available for Microsoft Windows when the bureau started development for the 2000 Census in the early 1990s.
CASES works like a paper questionnaire, Polzer said. The temporary enumerators fill in a code to identify themselves, and the software prompts them through the rest of the survey. Interviewing members of a typical household takes about 15 minutes.
The bureau also loaded the OmniBooks with CA-Clipper from Computer Associates International Inc. to provide a menu for selecting case management and telecommunications programs. The case management option shows the list of households scheduled for verification visits.
At the end of each day, the verification workers will dial into bureau servers using Connect:Remote software from Sterling Commerce Inc. of Dublin, Ohio. After passing several security checks, they will upload their results and download any assignments made since the last connection, Polzer said.
The bureau operates servers at two sites in the Washington area for redundancy, Polzer said. The 12 regional verification offices, in the same cities as the 12 regional Census offices [GCN, Feb. 7, Page 1], have Sun Microsystems Enterprise 5500 midrange servers running Oracle Corp. relational database software to make the assignments for enumerators.Down for the count
Once the counts are in from both the main census and the verification effort, bureau statisticians will mathematically analyze the data sets to arrive at a corrected estimate of the U.S. population as of April 1.
The corrected count can be used for all purposes except apportioning congressional seats. In January of last year, the Supreme Court ruled that the bureau's statistical sampling could not be used for that purpose.
The bureau has interpreted the ruling to mean that the corrected count is usable for all other purposes, Singh said.
'We will release two sets of numbers, one corrected and one uncorrected,' he said. Users can decide which set they want.
To keep the verification independent, Singh said, the main census and recount efforts have separate staffs and offices. The Census 2000 temporary workers can be hired for the verification effort, but they cannot go back to working on the main census and will not be assigned to the area they canvassed before.
Some of the clerical staff members at the bureau's National Processing Center in Jeffersonville, Ind., will be involved in the verification, but not those at the three other regional Census 2000 data capture centers.
Census has adopted CAPIs for some of its other surveys to speed up processing because keying or scanning of data is eliminated, Polzer said.
Computer-toting Census enumerators will not appear on every street in America, however. 'The thought of using laptops on the [decennial] census is just overwhelming,' she said.
Once the OmniBooks are returned, bureau officials will reformat some of them for an undetermined version of Windows to use in other surveys, Polzer said.