Every 10 years since 1790, US officials have fanned out across the nation to try to count each person residing in the country.
And, not surprisingly, every decade since 1790, they have failed at that task.
While census takers find and count more than 98 percent of Americans, they consistently are unable to locate a particularly elusive segment of the population that includes the rural poor, minorities, renters, recent immigrants, and children.
By some estimates, the uncounted number 4 million to 5 million.
After years of research, the United States Census Bureau decided to use statistical sampling methods to compensate for this chronic population undercount. But the bureau's plan for the 2000 census has sparked a political firestorm over whether federal law and the Constitution permit the use of sampling techniques.
Argument on both sides
Today, the US Supreme Court will hear arguments on both sides of the issue.
Opponents of sampling, including many Republicans in Congress, say the Founding Fathers mandated a physical head count to ascertain the population.
Sampling proponents, including the Clinton administration and many congressional Democrats, say statistical techniques can help produce a more accurate count and guarantee full representation for the entire population.
The issue is critical to members of Congress because the census is used to divide the 435 House of Representative seats and other legislative districts nationwide. In addition, it is of great interest to cities and counties because census population statistics determine the allocation of billions of dollars in federal aid.
If a majority of justices uphold sampling, the resulting slight population adjustments to the overall census data could mean more federal dollars spent in areas plagued by past undercounts, such as Los Angeles, New York City, Miami, and parts of Texas. And that could rewrite the map of congressional representation, with Indiana, Connecticut, Massachusetts, Minnesota, Missouri, Pennsylvania, and Wisconsin - among other states - possibly losing a seat in the House of Representatives.
Although the battle is being waged with legal and constitutional arguments, some analysts say the motivating force behind it is pure politics. These analysts say Republicans perceive the census undercount as helping them maintain control of Congress, and Democrats believe the recognition of previously uncounted Americans will strengthen their hand.
Is sampling scientifically reliable?
On a more technical level is the issue of whether sampling is scientifically reliable.
Among professional statisticians and demographers, statistical sampling is considered to be a valid and important tool. The American Statistical Association filed a brief in the case saying that "properly designed sampling is often a better and more accurate method of gaining knowledge than an inevitably incomplete attempt to survey all members of such a population."
Others don't see it that way. In a brief filed on behalf of the Foundation to Preserve the Integrity of the Census, James Hamlin writes, "the statistical sampling proposed by the bureau will almost certainly make the census less accurate and undermine its legitimacy."
Beyond the political and technical questions, the heart of the case before the high court is whether sampling is legal.
Two different federal courts earlier this year found that federal census law bars the Census Bureau from relying on sampling in a population count used to apportion seats in Congress. The law is ambiguous, with one section seeming to permit sampling and the other to prohibit it.
If the court determines that the law does not permit sampling, the case may end there. But if a majority of justices rule there is a legal provision for the use of statistical techniques, sampling opponents will argue that such methods nonetheless violate the Constitution.
The Constitution requires an "actual Enumeration," which Republicans argue means a head count. Democrats counter that the Constitution assigns to Congress the responsibility of determining how that enumeration is achieved.
One technical issue could end the case quickly. If the justices decide the suing parties can't show direct harm to themselves, the case will be thrown out without ever exploring the broader legality of sampling.
Opponents say they are concerned sampling might make the census vulnerable to political manipulation. They say census workers could create several different statistical models, each resulting in a different apportionment of congressional seats.
Statisticians and other experts say such manipulation would be nearly impossible to carry off without being detected by the Census Bureau staff and independent observers outside government.