News Release

Advance Notice of Article to Appear in the Notices

Contact At AMS: Tim Goggins (401) 455-4110, FAX (401) 331-3842, tjg@ams.org

Providence, RI -- In August of last year, a Federal Court ruled in a suit brought by the Speaker of the House of Representatives that statistical sampling violated the Census Act. This ruling paved the way for a prohibition against the use of statistical sampling in the 2000 census. The case has now reached the Supreme Court, which is expected to hand down a ruling in March. Politics has cast a shadow over the debate on this question, because census data is used to apportion seats for Congress.

``Estimation based on statistically designed samples is a standard and widely used method for obtaining information about large human populations,'' explains David S. Moore in the enclosed editorial, which will appear in the March 1999 issue of Notices of the American Mathematical Society. One of the nation's leading statisticians, Moore served last year as President of the American Statistical Association.

The basic idea behind statistical sampling is that a subset of a population can represent the characteristics of the population as a whole. Therefore, if one wants to collect information about a certain population, one need not survey every single individual. Gathering information from a randomly selected sample of individuals allows one to avoid biases and makes the sampling statistically trustworthy. One can apply to the information from the representative sample well established statistical principles to extrapolate information about the whole population.

Statisticians, mathematicians, and scientists regarded with surprise and dismay the prospect that the Census Bureau would be barred from using statistical sampling in the 2000 census. While legitimate scientific questions can be raised about the details of how the Bureau might best employ statistical sampling, it is clear that proper use of this tool can improve the accuracy of the census.

Concerned about the influence of politics on questions that are basically of a scientific nature, Moore warns that the ``distrust of national statistics may grow worse'' if politics influences the work of non-partisan groups like the Census Bureau. ``The fallout from oversimplified and sometimes irresponsible claims will damage statistics as a discipline, government statistical offices, and public trust in important national data,'' he writes.