Today a committee of senior U.S. Census officials will make a recommendation with $185 billion at stake. Their recommendation to Bush-appointed Census director William Barron concerns the use of statistical sampling for allocating federal aid. Statistical sampling has long been a source of partisan contention. Democrats, who would be more likely to gain funding for their districts if sampling were implemented, support the practice arguing that it is the only way to accurately count ethnic minorities, urban dwellers and the poor. Republicans, many of whom stand to lose funding for their home districts if sampling is used, oppose sampling on the grounds that it is a violation of the Constitution. Due to the inherent complications of conducting the Census, sampling is the only way to ensure a fair and equal count.

Traditionally the Census has been conducted as a raw head count. However, this massive logistical effort is marked by shortcomings. The system is expensive, costing over $4 billion, and rife with inaccuracy. It is believed that the 1990 Census did not count 8 million people, while double-counting 4 million individuals. The undercount disproportionately affected minorities, inhabitants of urban areas and the poor, with about three times the national average of minorities not being counted. Further compounding this problem is the fact that suburbanites are more likely to be counted twice. As a result of these gross inaccuracies federal funding is not justly distributed. Those with the greatest needs are simply ignored.

Statistical sampling is an effort to correct this problem. Using sophisticated techniques the raw head count is used for 90 percent of the population while the remaining 10 percent is determined by sampled data. This data is checked for accuracy by comparing it with a sample count of 314,000 households. It is estimated that sampling will decrease the Census" error rate from its present level of 1.2 percent to 0.1 percent. The American Sociological Association, National Academy of Sciences, American Statistical Association and the General Accounting Office all support the use of sampling.

Republicans cite Article 1, Section 2, Clause 3 of the Constitution in their argument against sampling. The language calls for the "actual enumeration" of individuals. Republicans contend that this requires a nation-wide head count. However, the Census has used forms of statistical sampling since 1940. In 1940 the Census first used the "long form" which is sent to only 1/6 of the population. The data from those who receive the "long form" is extrapolated to make generalizations about the rest of the nation. According to the Census" 1997 report to Congress, "Census takers have never been able to contact and count each and every resident of this nation. As a result, information on less than the whole population has always been used to characterize the whole population." At best Republican concerns over "actual enumeration" are misguided at worst they are a duplicitous attempt to deprive millions of justly deserved aid.

The question of statistical sampling is too important for partisan interests to decide. If statistical sampling is not used Detroit would lose $230 million over the next ten years because undercounts could put the city in a different population category. Science should be used to remedy the inherent flaws of the Census and more accurately represent all Americans.