(December 2000) Census 2000 is winding up. The U.S. Census Bureau will announce final population counts by year’s end and release additional tabulations starting in early 2002. The 2000 census includes many improvements, such as redesigned questionnaires, a broader publicity and outreach campaign, and new partnerships with state and local governments. By all accounts, response rates have exceeded expectations, and the Census Bureau has kept the process on or ahead of schedule. So why only two cheers for the census and not three? Because two major improvements that could have been incorporated in the 2000 census were not included.
First, the Census Bureau was not allowed to implement a plan to sample nonrespondents. The original plan for the 2000 census included a sample-based follow-up of households that did not mail back the questionnaire. Such a plan would have maintained an acceptable level of data quality while yielding major cost savings. But beginning in the mid-1990s, some congressional representatives and advocacy groups challenged the legality of the Census Bureau’s 2000 plans. Their lawsuits were combined and brought before the Supreme Court. The Court, relying on language in the Census Act (rules for the Census Bureau that Congress enacted and only Congress can change), ruled in January 1999 that the apportionment count from the census could not be based on sampling. Accordingly, sampling for nonresponse follow-up was not permitted.
The ruling carried a high price tag. Cost estimates made by the General Accounting Office (GAO) in the early 1990s indicated that, if trends in mail response rates continued, the 2000 census might cost as much as $4.8 billion, compared with the 1990 Census costs of $2.6 billion, in 1990 constant dollars. But, because of extra efforts to increase the mail response rates and the need to follow up all nonrespondents, 2000 census costs are estimated by Census Bureau Director Kenneth Prewitt to exceed the GAO’s projection by about $1.5 billion. Overall 2000 census costs will be $6.2 billion in current dollars, more than double the price of the 1990 Census. Legal and political opposition to the use of sample-based follow-up of nonrespondents has therefore resulted in the most expensive census to date.
Second, the Census Bureau was directed to conduct two different types of censuses at the same time in 2000. To achieve the most accurate direct population count possible, which will serve as the basis of the census apportionment numbers, census officials have made intensive and costly efforts to count every person. In addition, the Census Bureau is conducting an independent survey to estimate the undercount and to adjust the final census numbers. The 2000 census, in other words, includes a direct-count census and a census based on survey adjustment.
These two operations duplicate one another and increase overall costs. Once the decision was made to use a survey to adjust for the undercount, the entire census operation should have been reengineered to avoid duplication and achieve substantial cost savings. Estimates made in the mid-1990s in the National Academy of Sciences’ Modernizing the Census report suggested that census costs could be reduced by as much as $1 billion through such reengineering.
If the two “missing” improvements described earlier had been implemented, the potential cost savings in the 2000 census would have amounted to at least $2.5 billion ($1.5 billion by using sampling for nonresponse follow-up and $1 billion from reengineering for a census relying on a survey to estimate undercount), or more than 40 percent of the total $6.2 billion that the 2000 census likely will cost. Cost savings are the single most notable missing element in the 2000 census. The failure to take these two missing improvements into account was not the fault of the Census Bureau. Congress, in accordance with the Census Act, directed the Census Bureau to rely on two census designs in 2000. The two designs cause needless and expensive duplication.
Barry Edmonston directs the Population Research Center and is a professor of urban studies and planning at Portland State University in Oregon.