|
Science
Friday > Archives
> 1998
> August
> August 28, 1998:
Hour One: Census Sampling
| This week, a Federal court ruled that the Census Bureau would not be allowed to use statistical sampling techniques when it comes time for the next national count in the year 2000. It's a debate that has its basis in science, law, and politics. Pointing to language in the Census Act of 1976, the three-member panel of judges agreed with a group of Republican members of Congress who had brought suit against the Administration to stop the use of sampling. But the Census Bureau points to its need for efficient, accurate counting -- and says that sampling is the way to go. They plan to appeal the decision. |  | The last census, in 1990, is generally agreed to have had major problems. According to estimates by the Census Bureau, that census undercounted the population by 1.8 percent -- and, even more of a problem, the undercount was not even across the country. Minority and inner-city populations were especially undercounted, critics of the 1990 census claim. The census count is the basis for apportionment for congressional seats and for the distribution of some federal money to states, so inaccuracies in the count can be significant. 
How it works:
Step 1. In traditional census taking, a form is mailed to
every home thought to contain people. Steps 2-5.
Enumerators repeatedly visit every home that doesn't
respond by mail. Some people may go uncounted.
Steps 1-3. When sampling is used, traditional methods are
used until 90% of the population is counted. Step 4.
Information about the remaining 10% is extrapolated from
the data collected in other homes. |
The Census Bureau planned to use sampling to supplement the information they got from the traditional count, made using mail-in questionnaires and in-home visits by census takers. Many people never mail in the census questionnaire, and sending census takers on repeated visits to unresponsive homes can quickly get expensive. The Census Bureau wants to use traditional methods to count at least 90% of the population in any given area, and then draw conclusions about the remaining 10% using statistics. The plan is supported by statisticians from both the National Academy of Sciences and the American Statistical Association.
But the Constitution requires Congress to obtain an "actual enumeration" every ten years. Sampling, opponents protest, is not an "actual enumeration," could lead to more inaccuracies, and could be open to manipulation and political fraud.
More cynical observers point out that opponents of the plan may also be afraid that a census that more accurately counts minority and inner-city populations might help swing the balance of power towards the Democratic side -- explaining why many of the opponents of the plan tend to be Republicans.
The next stop for the legal debate is the Supreme Court. Join us on this hour of Science Friday for a look at the history, politics, and science of the census. |
You tell us:
Should the government be allowed to use
statistical sampling in the Year 2000 Census?
Here's what some listeners said!
Guests:
Stephen Fienberg
Maurice Falk University Professor of Statistics and Social Science
Carnegie Mellon University
Pittsburgh, PA
Steve Holmes
Correspondent
The New York Times
Washington, DC
Harvey M. Choldin
Author, "Looking for the Last Percent: The Controversy over Census Undercounts" (Rutgers University Press)
Professor Emeritus, Sociology
University of Illinois
Champaign-Urbana, IL
Books/Articles Discussed:
"Looking for the Last Percent: The Controversy over Census Undercounts" by Harvey M. Choldin. Rutgers University Press, 1994.
Related Links:
US Census Bureau - Census 2000
House Subcommittee on the Census
American Statistical Association
General Accounting Office reports on the Census
Southeastern Legal Foundation
|
