Sampling is a scientific procedure used in the social sciences to make more accurate generalizations about large groups. Measuring large groups, like the entire population of a country, can be too difficult because of cost, time or ability. Social scientists have developed the use of representative samples selected from larger populations in order to make estimations about important characteristics.

In 1990, the Census Bureau determined it had missed about 8.4 million people and double-counted 4.4 million others. The net undercount of 4 million represented 1.6% of the total population. Collecting accurate information, particularly from unresponsive households, has become increasingly difficult:

  • The rise in the number of two-earner households means that fewer people are home during the day when census enumerators visit.
  • Because of the increase in foreign-born citizens it has become more important to have more enumerators who speak a language other than English.
  • In gated communities or apartment buildings with locked entries, enumerators have trouble even reaching the households.
  • In some low-income areas, residents' high mistrust of government and census-takers' fear of crime might reduce the count's accuracy.
  • Children are more likely to be left off the census form because so many live in binuclear homes or stay with relatives.

Congress debated the Census Bureau's plans to use scientific sampling to supplement direct counting methods in the 2000 Census. The U.S. Supreme Court ruled 5-4 in January 1999 that the Census Act amendments bar the use of sampling methods to compile the state population totals used to apportion the 435 seats in Congress among the 50 states. Justice Sandra Day O'Connor wrote the majority opinion. In her opinion she found that the law allowed the use of sampling for purposes other than apportionment – if feasible. The Supreme Court also said that the Census Act amendments gave the Secretary of Commerce substantial authority in determining the best way to conduct the census – including the use of sampling procedures. The Census Bureau unveiled a revised Census 2000 plan in February 1999, to comply with the Supreme Court's interpretation of the law. In March 2001 a committee of twelve senior Census Bureau professionals recommended the release of unadjusted census numbers to the states for redistricting purposes.

Two forms were used for conducting the Census, a long form and a short form. Most people were asked to complete the short form. Each household received either a short form (100-percent questions) or a long form (100-percent and sample questions). The long form questionnaire included the same 6 population questions and 1 housing question that are on the short form, plus 26 additional population questions, and 20 additional housing questions.

Though not for the purposes of congressional apportionment, the Census Bureau used a sampling ratio of about one long form (sample) questionnaire for every six households to obtain sample data on content as it has in previous censuses. Sample questions included, place of birth, work status last year, income, ancestry, monthly rent, veteran status, disability, plumbing and kitchen facilities.

To learn more about the methodology of the 2000 Census go to these web sites:

Critical Thinking

Question 1: Economic decisions are often political decisions. In this case, why does the Democratic Party support the sampling method instead of the traditional way of conducting the census? Why does the Republican Party oppose the use of sampling methods?

Question 2: Conducting an exact count of the population is increasingly more difficult. What kinds of problems associated with conducting a census might we expect to continue in the future? What are some strategies that might increase the participation of everyone in the next census?

Student Name: 
Student e-mail: 

Faculty Name: 
Faculty e-mail: