Washington State’s population is growing fast enough that the state seems likely to warrant another Congressional seat (our tenth) when the U.S. Census is taken in April, 2000–only four years from now. But the outcome is sufficiently uncertain that it may depend on what kind of Census the government holds. Will Congress fund a traditional Census that counts each person? Or, bowing to the Administration, will it authorize a new kind of Census that enumerates most people, but then uses population sampling techniques to complete the job? On such an arcane issue hangs great electoral power and the public’s confidence in the statistical system whose numbers we rely upon everyday in commerce, education and politics.
It is true that sampling would eliminate many time consuming house calls and other enumeration efforts, and thus save hundreds of millions of dollars. But Census numbers are used to distribute large shares of the federal budget, as well as to apportion seats in the House of Representatives and redistrict state and local legislatures. So it is worth doing right. The “cheapest” Census, after all, would be one that didn’t even attempt a complete enumeration and simply came up with numbers based on surveys. Yet most of us would not be represented in those numbers as real people, but only as surrogates for certain demographic characteristics, such as age, race, sex, and income level.
It is also true that the Census misses several million people in its traditional count. But such an “undercount” has been experienced each decade for 200 years without endangering the republic. In 1990 the undercount amounted to only an estimated 1.8 percent. Instead of creating artificial numbers to represent the missed residents (including many who wished to be missed), the Census should emphasize improvements in taking the count. These include employing new software and using administrative records, such as Social Security and welfare rolls, to find persons omitted in the head count.
If, instead, we undertake an official census that mixes virtual people with real people what will happen to public trust in the results? What happens when the Census imputes such virtual people into neighborhoods and voting districts where critics can show, block by block, that the bodies just aren’t there?
The Administration professes to be worried about all the lawsuits that were filed after the 1980 and 1990 censuses by states that lost Congressional seats. Ironically, the government so far has won all those suits. But if the Administration cares about legal vulnerability, just wait until it gets hit with the suits brought by states that lose representation through a Census whose adjusted numbers can easily be disproven in small areas.
In Washington State, it is very possible that we would not have received our ninth Congressional seat in 1990 if the Census had been adjusted that year. It was a close call as it was. In 2000, despite an expected population growth of roughly one million new people–from 4.9 million to 5.9 million, a 20 percent growth rate that is twice that of the nation as a whole–the chance of a tenth seat for Washington could be another close call. Once again, that chance could be jeopardized by Census adjustment.
According to the Forecast Division of the Office of Financial Management in Olympia, the state’s population already has passed 5.5 million, an increase about the size of the average current Congressional district nationally, 569 thousand. However, each state’s prospects are affected by growth rates in other states. The overall U.S. population is expected to rise from 248 million in 1990 to about 275 million in 2000, so the average district population will have to grow, too. In Washington, that will mean an increase from districts of 540 thousand each to districts of about 632 thousand each.
No one knows how all this arithmetic will work out because no one knows whether current growth rates will continue for four more years. But what does seem certain is that states like Washington that do a relatively good job of counting their people will suffer if the Census Bureau awards other states with large imputed numbers. What also seems certain is that the Census Bureau’s hard-earned reputation for integrity could suffer when people find out that a mere survey has removed part of the foundation of our national statistical system. It would be analgous to announcing that we were going to settle future political contests with polls rather than actual elections.
If the Census Bureau’s statisticians are ordered to do so by the Administration, and Congress goes along, they will conduct surveys and make subsequent adjustments as fairly as possible; for they are a conscientious group. But eventually the traditional civic role of the Census–one of the few functions of government named in the Constitution–will be undermined and its political legitimacy brought into question. As a government demographer (who wishes anonymity) told me recently, “The statistical community is not united on this. There is a potential here for cooking the books that is stomach wrenching.”
Congress, therefore, should not go along. Why create yet another source of public mistrust?