The short form contained only seven questions, the shortest census questionnaire since 1820. The long form asked 52 questions of approximately 1 in 6 households (approximately a 17 percent sample of the population). In previous censuses, responses to the race question were limited to a single category; in 2000, for the first time, respondents could check as many boxes as necessary to identify their race. A 1996 law mandated a new question on grandparents as care givers. Questions on disability were expanded to include hearing and vision impairment and problems with learning, remembering or concentrating. Questions on children ever born, source of water, sewage disposal and condominium status, were dropped; the 1990 census short form question on rent and property value became a long form question.
To counter a decline in the questionnaire mail-back rate, the Census Bureau embarked on an aggressive paid advertising campaign, awarding a $167 million contract to the Young and Rubicam Company for national and local print, television and public advertising campaign. This campaign consisted of more than 250 television, print, radio, outdoor, and other advertisements in 17 languages; it reached 99 percent of all U.S. residents. By the end of the campaign, the census message - "This is your future. Don't leave it blank." - had been seen or heard an average of 50 times per person. This campaign, along with an aggressive non-response follow-up program, brought the mail-back rate up to about 67 percent.
There were additional options for responding to the census. People receiving the short form could respond on the Internet, and about 70,000 households did so. Telephone questionnaire assistance centers provided questionnaire help in 6 languages and took responses to the short form over the phone.
Following the costly litigation generated by the 1980 and 1990 censuses, particularly that which sought statistical adjustment of the census counts to correct for persons estimated to have been missed or duplicated, the Census Bureau designed a plan for the 2000 census that it believed would eliminate the possibility of litigation. The Census Bureau's May 1995 plan for a "reengineered census" was the culmination of a four-year process of discussion and review of census plans by a broad spectrum of experts, advisors, and stakeholders. The plan was further refined and on February 28, 1996, Commerce and Census Bureau officials made public.
The plan called for using statistical sampling techniques in two principal ways. The first was to alter the traditional 100-percent personal visit of non-responding households during nonresponse followup (NRFU). Instead, a small percentage of non-responding households would be followed up on a sample basis. Information from this sample would be used to estimate the number of persons and their characteristics in the remaining non-responding households.
The second involved the Census Bureau taking a sample of 750,000 housing units to be matched to housing unit questionnaires obtained from mail and telephone responses as well as personal visits. The goal of this quality check survey was to develop adjustment factors for persons estimated to have been missed or duplicated in the census and correct the census counts to produce one set of numbers. This was to be a "one-number census," corrected for net coverage errors, that is, Integrated Coverage Measurement (ICM). ICM was a significant departure from 1990, when the results of the post-enumeration survey (PES) were used to produce a separate set of statistically adjusted counts after the delivery of the apportionment counts and redistricting data. This resulted in two competing sets of population numbers.
Beginning with the fiscal year (FY) 1997 appropriations process, the congressional majority included language in appropriations legislation that would prohibit the use of sampling in Census 2000 or the expenditure of funds for Census 2000 sampling-related planning activities.
Debate over the sampling issue postponed passage of the Commerce Department's FY 1998 appropriations bill until the end of November 1997, two months into the new fiscal year. With the threat of a stalemate between the congressional leadership and the Clinton administration in the debate over the use of statistical sampling in Census 2000, the two sides reached a compromise in the enacted legislation. The legislation, among other things, permitted the Census Bureau to continue to plan for sampling, while directing the agency to plan for a census without statistical sampling. This was later referred to as "dual-track" planning.
The law also set up an eight-member "Census Monitoring Board" to observe and monitor all aspects of the planning and implementation of Census 2000.
Two lawsuits were filed in February 1998 – including one filed by the U.S. House of Representatives – that challenged the constitutionality and legality of the planned uses of sampling to produce the apportionment counts. In both cases, the federal district courts decided for the plaintiffs. The Department of Commerce sought review of the district court decisions by the United States Supreme Court. The Supreme Court agreed to hear the cases, and oral argument took place on November 30, 1998. On January 25, 1999, the Court held in Department of Commerce v. U.S. House of Representatives that the Census Bureau's proposed plan to use statistical sampling in the decennial census for purposes of determining congressional apportionment violates the Census Act (the Census Bureau's authorizing statute).
Thus, the Census Bureau could no longer pursue its Census 2000 plan that included the Integrated Coverage Measurement (ICM) program and sampling for nonresponse followup. However, following the Supreme Court ruling, the Census Bureau issued a revised plan in which it stated that it planned to produce statistically adjusted data for non-apportionment uses of census data, including redistricting. But in March 2001, following the delivery of the apportionment counts, the Census Bureau recommended against the use of the adjusted data for redistricting, because of concerns regarding their accuracy. The secretary of commerce accepted the recommendation and determined that the unadjusted data would be released as the Census Bureau's official redistricting data. In October 2001, after considering the possible use of the adjusted data for non-redistricting purposes – for example, their incorporation in the Census 2000 long-form (sample) data products – the Census Bureau director rejected the use of the adjusted data for such purposes.
The Internet became the principal dissemination medium for 2000 census data. All four of the detailed data files, now called Summary Files, were available to be downloaded as soon as they were released. Individual tables could be viewed through the Census Bureau's online database, known as the American FactFinder. Additionally these files were available for purchase on CD-ROM and DVD. The number of printed publications was reduced and the number of printed pages in the report series was by about one-tenth of the 1990 census publication program.