WASHINGTON _ Facing soaring costs and evidence that the 1990 census was in some ways the least accurate in decades, the Census Bureau is planning wholesale changes in how it will collect data in the year 2000 and beyond.
One change that is expected to be adopted, census officials say, is the use of sophisticated estimates based on surveys to supplement the actual counting _ a volatile issue that was the center of a furious partisan battle in the last census.
The bureau is also considering scrapping the long-form survey that has been used once a decade to gather information as varied as household incomes or how many telephones a particular residence has.
In its place, the bureau plans extensive monthly surveys conducted over an entire decade, providing a more timely flow of this broad demographic data.
To increase the percentage of households that mail back the census form _ which hit a record low of 65 percent in 1990 _ the bureau is also considering a number of new steps like dispensing census forms at post offices and other places and permitting people to respond by telephone to a toll-free 800 number.
The extent of the changes _ driven by steeply rising costs and the increasing diversity of the population _ has yet to be determined. In addition, Congress may decide to Limit the changes or keep things as they are.
The Census Bureau will test many of the ideas next year in New Haven, Conn., Paterson, N.J., Oakland, Calif., and six rural Louisiana parishes.
"These would be tremendous changes from the way we are doing things now," said Robert Tortora, the bureau's associate director for Statistics, Methodology and Standards, who is in charge of redesigning the census.
Perhaps the most revolutionary change envisioned is the proposal to extrapolate a final tally from samples of those who are counted and those who do not return the census questionnaires.
In 1990, in an expensive endeavor, field workers called enumerators conducted up to six visits to each of nearly 35 million households that had not returned the forms, to determine how many people lived there.
In the next census, visits would be made to only a percentage of these housing units and Census Bureau statisticians would extrapolate from this sample the total number of people in all nonresponding households. The precise percentage of households slated for follow-up visits would be determined by the tests planned next year. …