(February 1999) The decennial census is an essential ingredient in the American democracy. No other source provides as much comprehensive information about who we are. In addition, no other data have such important consequences for the way we govern ourselves. Our representative government—and a huge amount of federal dollars—are distributed based on the census reports of the number of people living in different places. The 2000 census will be as important as any in our history.

Unfortunately, the census has become too expensive, too burdensome, and too inaccurate for the U.S. Census Bureau to conduct the 2000 census they way it conducted the 1990 census.

Redesign is inevitable. Congressional leaders of both political parties have agreed on one fact: The 2000 census must be different from previous censuses. But while there is consensus on the need for change, there is disagreement on the specific changes that should be implemented in the 2000 census.

Although the census count of the population of the United States has never been quite complete (no census ever is), public concerns about its incompleteness have increased in recent decades.

After the 1990 census, one which was conducted amidst the difficulties of counting a large and increasingly diverse and mobile population, two issues emerged. The 1990 census cost more than any other census, even after allowing for population growth, inflation, and declines in mail response rates; and the 1990 census didn't count some population groups as completely as it counted other groups—a lingering and growing problem also experienced in the censuses that preceded it. Given these tow problems, many experts began to question whether the census costs might continue to climb with no likelihood of narrowing the undercount.

Yet while these problems are discusses, and given that one possible solution to the undercount—sampling—has been removed from consideration for purposes of apportioning congressional seats among the states by a recent U.S. Supreme Court decision, the Census Bureau is in the final stages of preparation for the 2000 census. In April, the Census Bureau will begin printing 300 million questionnaires in order to have them ready for the start of the census in mid-March 2000.

During 1999, the Census Bureau will complete preparations of the nation's mailing address list. The bureau will send copies of mailing addresses to 39,000 local governments by the end of 1999, asking for their help in checking the accuracy of addresses and related census maps.

The Census Bureau will add about 5,000 temporary employees in 1999 and open hundreds of local census offices. These local offices will recruit and train at least 250,000 enumerators who will conduct special operations like enumerating the homeless, will contact nonresponding household, and will work on the large independent data-quality survey.

By the time census field operations are completes in mid-2000, data processing will already be in full swing. Data checking and tabulation will be continued—with the possible use of surveys to complete the count—in late 2000. The Census Bureau must report the official population counts to the president and to the secretary of commerce on Dec. 31, 2000.

This report discusses the serious and complex problems associated with taking a modern census—concentrating on what has been learned over the past eight years about options for reforming the traditional census.

Barry Edmonston directs Oregon's Center for Population Research and Census at Portland State University.