Proofiness - Charles Seife [75]
The case went all the way up to the Supreme Court. The American Statistical Association filed a brief with the court that made the case pretty clear: “Properly designed sampling is often a better and more accurate method of gaining such knowledge than an inevitably incomplete attempt to survey all members of such a population. . . . There are no sound scientific grounds for rejecting all use of statistical sampling in the 2000 census.” But the Supreme Court disagreed. In a five-to-four decision—the five most conservative judges versus the four most liberal—the Court determined that sampling was illegal. The apportionment of House seats, by law, had to be based upon flawed, highly error-prone population numbers that undercounted minority voters.
Even though the ruling evaded the question about whether the use of statistics ran contrary to the Constitution—whether the “actual enumeration” clause referred to a head count and nothing else—there’s no question that the conservative majority was hostile to the whole concept of sampling. In a concurring opinion penned by Antonin Scalia, one of the most conservative justices of the Court, Scalia strongly implied that any use of statistical techniques would make the founding fathers spin in their graves.
Scalia turned to eighteenth-century dictionaries to show that the phrase “actual enumeration” had to mean counting each individual person, one by one; he cited the 1773 Samuel Johnson dictionary, for example, which defined “enumerate” as “To reckon up singly; to count over distinctly; to number.” Aha! To reckon up singly! Scalia pounced: by using the term “enumerate,” the founding fathers meant to count each person, one by one. Thus the census “requires an actual counting, and not just an estimation of number.” Further, using statistical techniques will “give the party controlling Congress the power to distort representation in its own favor.” Only head counts, as inaccurate as they are, are free from manipulation. Thus sampling is unconstitutional.
This is a specious argument on several counts. First, even if Johnson’s dictionary was the key to the founding fathers’ intent, the definition “to number” is just as valid as “to reckon singly.” Besides, the word “enumeration” was merely an accident. It was inserted during the drafting of the U.S. Constitution by the Committee on Style, a group that made minor, nonmeaningful changes to the document to correct grammar and clarity. They changed the word “census” to “enumeration” for reasons unknown.75 So to dwell upon the finer points of the dictionary definition of “enumeration” is ignoring the fact that the founding fathers called for a “census,” and that “enumeration” was substituted for reasons of style.
More important, there’s no bright-line distinction between “actual counting” and “estimation.” As we’ve seen, counting is a measurement like any other, and is thus subject to error. This error is unavoidable; it turns even the best count into nothing more than an estimate, an approximation of the truth. And if there are other measurement techniques that give you a better approximation of the truth—such as statistical sampling—they deserve the title of “actual” more than counting does. Pretending otherwise is to place too much faith in the error-prone numbers that come from a head count: it is an act of disestimation.
Finally, the idea that head counts are free from manipulation is wrong. In fact, even