SAN JOSE, Calif. -- The Census Bureau is testing a plan to cut the U.S. census down to postcard size and -- to fill the resulting information gap -- create a huge federal electronic information system linking records from sources as varied as the Internal Revenue Service, county assessors and direct-mail marketers.
Faced with pressure to save money, improve the count of minorities and enter the 21st century, census officials have been quietly soliciting opinions from statisticians and demographers on the possibility of a scaled-down census. It would be supplemented either by vast demographic surveys or by the linkup of existing electronic records, from the local to the federal level.
"These next few years, we are going to be looking at some things that people think are very radical," said Susan Miskura, chief of research and development for the bureau's Year 2000 Census.
Such "radical and fundamental change," as Ms. Miskura put it, is being pursued in the wake of the 1990 census, which set several unhappy records: It was the most costly census ever (at $2.6 billion), it elicited the least cooperation from the public, and it appears to have missed the most people -- particularly minorities.
The bureau, Ms. Miskura says, expects the new proposals to provoke "wide public debate."
None more so than the idea of a big, electronic federal database.
Marc Rotenberg, a staff attorney for Computer Professionals for Social Responsibility in Washington, said the census had an excellent record of protecting the confidentiality of its respondents. But -- despite Ms. Miskura's promise that "if we could not provide that protection, we would not do it" -- he does not believe that a central database could stay confidential. "I certainly wouldn't hook up with the Direct Marketing Association if I was interested in protecting privacy," he added. "That's a little bit like going to an arsonist to help select fire extinguishers."
Linda Gage, the California state demographer, said that linking existing records "is very difficult to do, technically." But even more difficult, Ms. Gage predicted, would be overcoming public fears of "a big database in the sky, a big centralized federal information system where someone can learn all about you."
"Although we may feel very confident that these records will be confidential, I don't think that will be the reaction of the public," she said.
Among statisticians and social scientists, the prospect of substituting surveys for most of the census can be just as controversial.
Although surveys may be less costly and easier than a census, they are much less accurate. And the special beauty of the census, its ability to describe communities and even neighborhoods and blocks with a fair degree of accuracy, would be lost if surveys were substituted.
In addition, Ms. Miskura said, census legal advisers are unsure whether surveys could legally fill the constitutional requirement for an "enumeration."
The Census Bureau will test several versions of a postcard-sized census next spring, said Robert M. Groves, an associate director of the census.
Tests, market research and focus groups will be used to learn whether a "significantly greater" number of Americans would cooperate with a shorter form, said Ms. Miskura. It would include only questions on age, sex, race and Hispanic origin.
Last year, people returned only 63 percent of the questionnaires the census mailed out, a drop of 20 percentage points from 1980.
Everyone received a short but confusing questionnaire with seven questions. And one in eight households was asked an additional 59 questions.
So, if the test shows that cooperation could be greatly improved, Ms. Miskura said, "What that would say to us is that we can do the census quicker and cheaper by using a short questionnaire, that it makes sense to look at gathering data that have been traditionally got by the census through other means."