Census relies on IT for faster, low-cost count

Faced with the task of counting hundreds of millions of people and delivering those results in just months, the Census Bureau last month began a dress rehearsal for the 2000 census by testing computers ranging from digital imaging systems to an online system that will disseminate customized census data.

Speed and the cost of the census are at the top of the list of concerns in Congress, particularly because the bureau missed counting about 4 million people in the 1990 census, which cost about $2.5 billion to complete. The bureau is required by law to deliver its results to the president by Dec. 31, 2000. In the last census, the bureau delivered initial results by the deadline but followed up with another set of numbers months later.

The bureau is looking to information technology to solve some of these problems. "Technology allows us to get the mail processed and the data captured faster so [that] in only eight months' time we can get the right and final census numbers delivered to the president," said Mike Longini, chief of the Decennial Systems and Contracts Management Office at Census. "Without technology, I don't think we could do it."

"Anything that can speed up the process and lower costs will be looked at very kindly on Capitol Hill," said Edward Spar, executive director of the Council of Professional Associations on Federal Statistics. "Congress is looking very closely at numbers this year."

Census numbers are used to set policy, distribute federal funds and apportion seats in the House of Representatives.

To help keep costs down, Census is relying on outsourcing more than it has previously— something Spar called a "wise decision" on the bureau's part.

For example, Lockheed Martin Corp. and its team are building imaging systems under the Data Capture System 2000 (DCS 2000) contract. In the 2000 census, the systems will digitally capture 1.5 billion pages of census forms and convert them into a computer-readable format.

After one week of scanning returned census forms in the dress rehearsal, Lockheed Martin has found that about 80 percent of the fields on the forms can be read by DCS 2000, said Clyde Relick, deputy program manager at Lockheed Martin. A higher recognition rate means the bureau will have to hire fewer people to manually key in the information.

"We are certainly looking at driving the cost down through technology and through the reduction of the number of people that have to be involved in the process," Relick said. "The technology itself— optical character and optical mark recognition— is fairly sophisticated. The error rate we would get is better than we would get from a single person keying in the information."

Another new system, being developed by a team led by IBM Corp., will help disseminate the census data via the Internet to the public. The Data Access and Dissemination System will give the public, for the first time, the ability to create customized reports and maps online using a combination of census data. The final version of DADS will be ready at the beginning of next year.

"Because this will be Web-based access to census data, we want to design a system that will be easy to use," said Mark Nelson, communications manager at IBM Global Government Industry. "Many people in the past would not think of using census data, but [now it] will be easy."

Census also plans to buy about 10,000 laptop computers for enumerators, who will visit 750,000 houses to check the accuracy of information on returned census forms. In addition, the bureau will award a contract later this year to automate about 520 remote local census offices and link them to the 12 Regional Census Centers, which manage field operations. Because the local offices will be based on client/server architecture, the bureau will get updated information right away, instead of a day or two later, as in 1990.

"The census is so big [that] we need to make decisions quickly," said Ed Wagner, a program manager in the Decennial Systems and Contracts Management Office at Census. "That's why getting the data on an hourly and daily basis is so important."