After the substantial undercounts of the 1990 census, when many Americans lost their share of representation in Congress, statisticians at the Census Bureau proposed the use of statistical sampling to find those Americans who were missed by traditional methods. The essence of the plan would have supplemented the traditional census with a high-intensity survey of randomly selected districts, so as to find the degree of undercounting in those areas; the original census would then be adjusted to reflect the better data. Although there were reasonable arguments challenging the effectiveness of specific sampling procedures, the Census Bureau felt that that sampling would provide a better chance of accuracy than does the present system.
But in the furor over sampling in the late 1990s, reasonable arguments were few and far between. After all, there are no Democratic or Republican statistical techniques--there are only Democratic and Republican voters, and the immigrants and urban poor who are frequently missed by such surveys generally fall in the former category. As could be expected, the debate turned on whose ox was gored; in 1991, Newt Gingrich wrote a letter urging the use of sampling to correct an undercount in Georgia, but six years later, as the GOP recognized the potential of sampling to increase Democratic influence, he wrote a similar letter opposing it.
The legitimate technical arguments on both sides of the sampling debate fell by the wayside, as opponents described the process as unconstitutional wild guessing--and supporters, upholding the counterintuitive position that guesses can improve accuracy, were hard pressed to explain standard deviations in sound bites. In the end, the matter was put to an end by a five-to-four Supreme Court decision, and the unadjusted numbers were released by the Census Bureau two weeks ago--but as with another recent five-to-four decision, the question has hardly died away.
Public confidence in government cannot but be shaken by spiteful conflict on an issue that seems so dry and so far removed from political influence. Unfortunately, the sampling battle demonstrated a principle that Americans have witnessed too often in the past election season: what seems to be merely technical can have immense political consequences. Ballot designs are apolitical in theory, but not so in practice; just ask the residents of Palm Beach County. The accuracy of recounts by hand as opposed to by machine may seem to be a simple comparison, but in November, hand counts were apparently good enough for Texas but unacceptable in Florida.
The reason why the election conflict brought so much anguish was not that feelings ran deep--they always do, in every election--but that the American political system was forced to resolve both the technical and the political questions at the same time. For the Florida Supreme Court examining "hypertechnical" recount requirements and for the American people examining the electoral college process, no apolitical answer existed to the apolitical questions. Everything from the mechanics of chads to the machinations of lawyers--every hard-and-fast judgment, indeed every number, became a subject for political debate, culminating in the public's introduction to Supreme Court arithmetic, where the only reasoning one needs to know is that five is greater than four.
Without a consensus on the technical details and on the questions reasonable people should be able to resolve (the population of the U.S., the number of votes cast for each candidate), one can hardly expect a civil process of political debate--for when reasonable people disagree and when reason fails as a means of persuasion, what is left but force? The past election was so painful because there was no comfortable realm of objectivity and unbiased judgment from which one could summon widely acceptable solutions to our disputes.
The sobering lesson to be drawn from the past few months is that the procedural is always substantive, that technical debates will always be resolved by a body politic manifestly unqualified to decide them. The issue is not that people are too stupid to govern themselves, but that no one, smart and educated people included, has the time to examine the issues in sufficient detail. How many citizens can be expected to learn statistics to understand the reasons for sampling? Is it unsurprising that the political and statistical debates are conducted on entirely different grounds?
Asking the people to decide only the ends to be achieved and leaving the means for achieving them to technically skilled bureaucrats offers no escape from this dilemma; for who will supervise the bureaucrats to safeguard the public interest, and who then will watch the watchers? Eventually, the people must be trusted to decide means as well as ends--and the technical and the political will be hopelessly entangled.