House Approves Ban on Sampling Technique in the 2000 Census

By Larry Jones

ljones@usmayors.org

Setting the stage for a possible showdown with the President, the House approved the 1998 appropriations bill for Commerce, Justice, State and the Judiciary (HR 2267) on September 30 by a vote of 227-199. The bill includes a controversial provision that would ban the U.S. Census Bureau from using statistical sampling until the Supreme Court rules on whether or not it is constitutional. The President vetoed a bill earlier this year because it included language banning the use of sampling, a statistical technique that the Bureau plans to use to account for individuals who do not respond to the census. President Bill Clinton has again vowed that he will veto this bill if it is sent to him with the ban on sampling. The Senate version of the bill (S.1022) which passed on July 29 does not include the ban on sampling, although the Republican leadership has expressed strong opposition to sampling. A conference committee composed of members from both chambers must now get together and resolve the differences between the two bills and reach agreement on a final compromise that must be approve by both chambers before it is sent to the President.

According to the Congressional Research Service (CRS), it is unlikely that the Supreme Court will issue a decision on sampling any time soon. In response to an inquiry from Representative Carolyn B. Maloney (N.Y.), CRS said that any decision on the constitutionality of sampling would most likely have to wait until someone suffers an actual injury as a result of the sampling technique. Democrats have charged that requiring judicial review of sampling amounts to a delay tactic which could tie the issue up in court for a year or more and prevent the use of sampling in the 2000 census count. The Census Bureau also stated that a delay would prevent the Bureau from carrying out its scheduled "dress rehearsal" next March to test sampling and all of the components of its plan for conducting the 2000 census in several cities and counties. The plan calls for testing new easier-to-read forms, multiple mailings, paid advertising, sampling individuals who do not respond by mail or telephone and quality check procedures.

Congress approved a ban on sampling in the supplemental appropriations bill (HR 1871) earlier this year but was later forced to drop the controversial provision after the President vetoed the bill. To avoid another fight with the President, the Senate did not include the ban on sampling in its version of the 1998 appropriations bill but opted instead for language that instructs the Census Bureau to make no irreversible plans to use sampling. During floor considerations of the House bill, Representatives Alan Mollahan (WV) and Christopher Shays (CT) offered a bipartisan amendment that would have dropped the ban on sampling and adopted language similar to the Senate bill. The amendment was defeated by a vote of 197-228.

Members of Congress are mostly divided along party lines on the sampling issue. Republicans, who mostly oppose sampling, argue that the constitution requires an actual enumeration and that sampling would be unconstitutional since it is used to estimate the number and characteristics of individuals in households that do not respond to the census. Democrats, who mostly favor sampling, argue that sampling is needed to account for individuals who do not respond to the census. They cite a number of studies that endorse the use of sampling as the most efficient way of ensuring an accurate census count. Without it, they argue that millions of people will be missed. According to Shays, who is a Republican member who supports sampling, Republicans are concerned that sampling may lead to more minorities and other individuals being accounted for who traditionally tend to vote Democratic. If this should happen, Republicans fear they could lose as many as 20 seats when House districts are reapportioned after the 2000 census.

Under the Bureau's plan, at least ninety percent of the households in every census track will be counted using the traditional method: mail, telephone and follow-up site visits from census workers. For the remaining ten percent who do not respond, the Bureau plans to collect information from a random sample and used the data to extrapolate the number and characteristics for all non-responding households. By using the sampling technique, the Bureau hopes to avoid the undercount that occurred during the 1990 census which was estimated at 4 million. If sampling is used in the 2000 census, Bureau officials estimate that the error rate will be 0.1 percent and that fewer than 300,000 people will be missed nation-wide. If it is not used, they estimates it will need an additional $800 million to hire more census counters to go door-to-door. Even with these additional workers, they predict the error rate will jump significantly to 1.9 percent, which would mean that 5.5 million people would be missed. Since the allocation of federal resources is based on census data, many areas stand to lose huge sums if sampling is not used.