08 Sep 1998 The 2000 Census: Sampling or a Straight Count?
Overview
When the federal government undertakes its official count of the population of the United States in the year 2000, the Clinton Administration believes a process of estimation called “sampling” is best way to achieve an accurate count. Congressional Republicans disagree, favoring the traditional head count used in past censuses. Although sampling has already been ruled illegal by a federal court, the Clinton Administration is still proceeding with plans to use it as it appeals its case to the U.S. Supreme Court.
What is the Census and Why Do We Need One?
The U.S. Constitution mandates a regular tally of the nation’s population. The means of accomplishing this national head count is called the census. The U.S. Census Bureau, an agency of the U.S. Department of Commerce, conducts the census. The Constitution requires a census be conducted every ten years.
Census data is important because it is used to determine the number of congressional districts awarded to each state. Likewise, data is used to create voting districts at the federal, state and local level. Census data is also used to calculate the allocation of federal money to various federal, state and local government programs throughout the nation.
How has the Census Been Conducted in the Past?
In past censuses, surveys have been mailed to every American household. If a household fails to return their completed survey, a representative of the Census Bureau would attempt to obtain the necessary information through an on-site visit to that household. This form of one-by-one head counting is called enumeration.
Because information could not be obtained from some households through both surveys and on-site visits, the results of the 1990 census are considered to only be 98.4% accurate. The Census Bureau estimates approximately four million people – many believed to be located in poor and minority communities – were not counted in 1990.
How is Sampling Different from Enumeration?
Under the Clinton Administration’s plan to employ statistical sampling for the 2000 census, the traditional enumeration process would be used to collect data from approximately 90% of the population. The remaining 10% of the population, approximately 26 million people, would be found through estimation determined by computer-generated statistical models. The “virtual people” created by this modeling would be derived from data collected during the on-site visit portion of the enumeration process. After going through a statistical manipulation called “integrated coverage management,” the figure would again be re-assessed in relation to super-accurate enumeration reports from selected areas to arrive at final figures for population and demographics.
Critics caution that relying on data drawn from a pool of people who did not originally comply with the government’s request for information would not reflect the overall demographic composition of an area.
Is Sampling Constitutional?
The U.S. Constitution states “the actual Enumeration shall be made… in such Manner as [Congress] shall by Law direct.” (Article 1, Section 2) The Census Act of 1976 specifically prohibits the use of statistical sampling in the creation of congressional districts.
While the Clinton Administration and congressional Democrats strongly favor using statistical sampling for the 2000 census, there are two ongoing legal challenges to the sampling process:
* Republicans in the U.S. House of Representatives, led by House Speaker Newt Gingrich (R-GA), filed a lawsuit against the Commerce Department challenging the use of statistical sampling in the 2000 census. In federal court, a lawyer representing the House Republicans pointed out that dictionaries published at the time the Constitution was written define enumeration as “counting one by one,” and that even modern dictionaries define the term as “to count” and “to name one by one.” On August 24, 1998, the specially-appointed judicial panel created to hear this case unanimously ruled sampling to be illegal under the terms of the Census Act of 1976, and issued an injunction against the use of sampling in the 2000 census. The Clinton Administration has said it will appeal the ruling to the U.S. Supreme Court, and is proceeding with a census plan that includes sampling. Likely Supreme Court review is expected to answer additional questions about the constitutionality of sampling that remain unanswered by the lower court’s ruling.
* The Southeastern Legal Foundation (SLF) filed a lawsuit against President Bill Clinton, the Commerce Department and Commerce Secretary William Daley and the Census Bureau and Acting Director James Holmes on behalf of 16 individuals, one county and U.S. Representative Bob Barr (R-GA). This lawsuit challenges the constitutionality of statistical sampling in the year 2000 census process. According to SLF President Matthew J. Glavin, “Statistical sampling is specifically prohibited by the Census Act, and the U.S. Constitution mandates an ‘actual enumeration’ by ‘counting the whole number of persons in each state’ to apportion Congress.” This case was heard in U.S. District Court in Roanoke, Virginia on August 4, 1998, and any ruling is also expected to be appealed to the U.S. Supreme Court.
Has Congress Approved the Use of Sampling in the 2000 Census?
On August 6, 1998, by a vote of 227 to 201, the U.S. House of Representatives passed a 1999 spending bill for the Commerce, State and Justice Departments that gives the Census Bureau access to only half of its $952 million appropriation. This would allow the agency to operate until March of 1999, at which time the Clinton Administration would have to seek congressional approval for the release of the remainder of the funds. Republicans say this restriction is necessary to monitor the accuracy and progress of census implementation. By March, final rulings in both lawsuits challenging statistical sampling should be available as well as a report by the bipartisan U.S. Census Monitoring Board.
Congressional Democrats charge Republicans are playing politics with the census, and President Clinton has threatened to veto the bill unless everything covered by the spending bill is limited to only six months or the Census Bureau receives full access to its appropriation. A veto would deny funding to the Commerce, State and Justice Departments, causing a partial government shutdown. Last year, Clinton vetoed a disaster-relief bill because it contained an amendment banning the use of sampling. At the time, Republican opponents of sampling backed down.
U.S. Representative Dan Miller (R-FL), a former statistics professor who is opposed to using sampling in the 2000 census, is quoted by The Washington Times asking, “Why does Mr. Clinton want to shut down the FBI and take cops off our streets over the 2000 census?” Miller is also planning to introduce legislation to prohibit census procedures from deleting “real people” from any population figures derived from sampling unless these persons are erroneously double-counted.
What are the Concerns about Sampling?
Critics of sampling argue that:
* Sampling is unreliable. A report by the Census Bureau found that sampling is less accurate than enumeration in cities with populations of 100,000 or less. According to 1994 Census Bureau figures, 58.6% of the population lives in “incorporated places” of less than 100,000 people. Tests involving sampling conducted during the 1990 census would have mistakenly taken away a congressional seat from Pennsylvania in favor of Arizona.
* Real people may be replaced by virtual people. Sampling could remove people who legitimately filled out census surveys in order to obtain demographic outcomes consistent with statistically driven interpretations of data.
* A flawed census could be costly in both time and money. The 2000 census is expected to cost approximately $4.2 billion. If the results are successfully challenged in court, it could invalidate all census data and necessitate a recount. Advocates of sampling say the sampling process could save $100 million. At the risk of a costly repeat count that would leave proper congressional representation and federal funding mechanisms unresolved for years, enumeration proponents shun the notion that putting short-term savings over accuracy will save money over the long term.
* Sampling is polling, and polling is not accurate. President Clinton alluded to the guesswork involved in sampling when he compared the process to polling during a June 2, 1998 speech in Houston, Texas: “Most people understand that a poll taken before an election is a statistical sample… And sometimes it’s wrong, but more often than not it’s right.” Putting this statement into practice, a memo from the Republican activist organization GOPAC pointed out: “In 1996, most of the national polling data showed that [Clinton] would beat Senator Bob Dole by 20 points. The last pre-election poll – conducted by CBS – put the Clinton-Gore ticket ahead by 16 points. That’s about 15 million out of 94,683,948 votes cast. Well, [Clinton] won, but by 8%, not 16%. What happened to the other 7.5 million votes? Should we let 7.5 million people drop out of the population through a sampling error? (that’s equivalent to the population of 12 congressional seats, by the way).”
* Sampling could be influenced by partisan politics. An Atlanta Journal and Constitution editorial on June 15, 1998 stated: “We object to sampling because, frankly, we just don’t think a highly partisan administration can be trusted to be coldly honest when dealing with a process that, by its nature, involves interpretation, estimation and manipulation of figures. The temptation to ‘help’ the results along would simply be too great – whether the administration was Democratic or Republican.” The Clinton Administration has already been accused of misusing federal agencies like the Federal Bureau of Investigation, Internal Revenue Service, Immigration and Naturalization Service and even the Department of Commerce (which is in charge of the Census Bureau), among others, for political gain.
* Sampling admits defeat before the counting process even begins. The 1990 census is estimated to have missed less that 2% of the population, yet sampling intentionally omits 10% of the population. Instead of doing the best possible job of counting by enumeration and sampling from that point, the Clinton Administration’s sampling plan raises the potential for error and abuse by expanding the percentage of the population that is estimated by sampling.
Are There Alternatives to Sampling?
J. Kenneth Blackwell, Ohio State Treasurer and co-chairman of the U.S. Census Monitoring Board, says that one way to count hard-to-find individuals and families – especially those in poor areas – is to use administrative records related to programs like Medicare and Social Security. Blackwell pointed out in a July 12, 1998 commentary in The Washington Times that the 1990 census added 1.3 million people through parole and probation records alone. “The method,” Blackwell wrote, “adds people frequently missed in a conventional enumeration, without subtracting people who were not.”
Republican leaders in Congress also pledged to provide as much money as the Census Bureau needs to provide an accurate enumeration for the 2000 census. The Washington Times quoted an unnamed Republican leader in Congress who pledged himself and his colleagues to “providing a level of funding that is sufficient to perform the entire range of constitutional census activities with a particular emphasis [on counting] all groups that have historically been undercounted.” Susan Irby, a spokesperson for Senate Majority Leader Trent Lott (R-MS), added, “We would be more willing to spend more to get an accurate count, rather than getting a political count using the political sampling the Administration is proposing.”
David W. Almasi is Director of Publications and Media Relations of The National Center for Public Policy Research. Comments may be sent to [email protected].