|Census 2000: Another Technological Revolution?|
(Technical and general audiences)
In 1890, it was keypunch cards; in 1950, UNIVAC I, the first civilian-use mainframe electronic computer, made its debut; for Census 2000, the Census Bureau could introduce yet another technological revolution.
Census 2000 will be the first fully computerized census, from collecting data to releasing the final results on the Internet. And, in the level of access to the data, it could be considered the most democratic of the 21 censuses of population and housing ever. Internet will give millions of more data users the tools they need to construct a digital picture of America down to any one of 7 million census blocks.
For the first time, the data capture process will use optical scanners that can read hand-printing to process the millions of questionnaires returned by mail or filled out by enumerators. Scanners, of course, have been used for decades to recognize the marks made on standardized I.Q. and college-admission tests. But the strokes making the marks were restricted to defined ovals or boxes and the markings themselves to No. 2 lead pencils.
By contrast, the optical scanners to be used in Census 2000 will recognize and decipher hand-written responses made by pens, as well as pencils. But the scanners are just one component of the system being developed by Lockheed-Martin Mission Systems. Lockheed is using commercially available sorters, scanners and processors -- rather than developing expensive new hardware and software -- in a system called Data Capture System 2000 (DCS 2000). Within this system, the scanners will take electronic photos, or images, of the census forms. Then, the "photos" will be processed by software capable of recognizing an infinite variety of hand strokes as either alphabetic or numeric characters. After the characters are translated into computer code, the responses will be transmitted electronically to the Census Bureau headquarters complex for statistical processing and analysis.
More than 100 million questionnaires will flow through the DCS 2000 systems at four data capture centers located around the country. Three of these centers, one in Baltimore County, Md., a second in Phoenix, Ariz., and the third in Pomona, Calif., are being established under a contract with TRW Inc. just for Census 2000. A fourth, in Jeffersonville, Ind., is a permanent installation called the National Processing Center, run directly by the Census Bureau to process its other censuses and surveys.
"Each center will employ between 1,500 and 2,000 people at the height of the census between March and July, 2000," according to Ann Gwynn of the Census Bureau's Data Capture Program. "TRW will be responsible for leasing, designing and building out space for the facilities, as well as hiring and training the workers who will operate DCS 2000."
As the forms come into the data capture centers, they will be checked in and the envelopes slit open for manual removal of the forms. The check-in process also captures information about which households have responded to the census. Those that do not will be contacted by a follow-up census-taker. Once the check-in function is completed, forms are fed into the digital scanners. The resulting images are sent to sophisticated computer processors where they are assessed for quality. Only then are the data "read" from the form. The goal is to minimize expensive manual keying operations and process the forms faster -- as much as 20 times faster.
As in the most recent censuses, two basic forms will be used, the so-called short- and long-form questionnaires. The short form, sent to roughly 83 percent of the estimated 118 million housing units in the country, asks for information on age, sex, race, Hispanic origin, household relationship and whether the housing unit is owned or rented. The long form seeks more detailed socio-economic information. The printing contract for the census questionnaires was scheduled to be awarded in November of 1998.
Census population totals are used to reapportion among the states the seats in the U.S. House of Representatives and, within states, to redraw the boundaries of state legislative districts and even some local voting districts. These totals and subsequent population estimates based on the decennial numbers are one element that determines eligibility or distribution to state and local governments of more than $180 billion a year in federal program funds (FY 1996 figures).
This census, moreover, will empower citizens to participate more directly than ever in political processes. Using the data disseminated by the Census Bureau via Internet, literally millions of people with access to personal computers will better understand the reasons behind zoning, road-building, new school or medical-facility decisions, as well as broader issues such as demographic trends and economic opportunities.
The use of keypunch cards in the 1890 census ushered in an era of business machines, tabulators, cash registers, etc. The computer age flowed out of UNIVAC I, used to process 1950 census data. With Census 2000 making it possible for anyone to log on to the same data used by government officials and interest groups, America will move closer to becoming an Internet Society -- a perpetual town-meeting in cyberspace.