Census Count May Be Low by 4.7 Million : Population: Undercount is far greater than that of 1980, estimates indicate. Usually, such errors are highest among blacks.

The 1990 census, already fraught with controversy, may have failed to count as many as 4.7 million U.S. residents, a major discrepancy between final population totals released this week and Census Bureau estimates issued last October.

The margin of difference could change as Census Bureau officials revise their estimates. But for now, the discrepancy offers the first statistical evidence that the census missed a lot more people in 1990 than it did in 1980--a development that critics of the census have been predicting all year.

Results of the 1990 census put the population of U.S. residents--not counting about 920,000 overseas federal workers--at 248.7 million. The figure is almost 3 million higher than the Census Bureau’s preliminary head count, announced earlier this year. But the final tally found 4.7 million fewer people than the October population estimate.

The potential undercount represents a 47% increase over the number of people missed in 1980, according to census figures. There is no information available yet indicating precisely who was missed in 1990 or where the omissions occurred.

Traditionally, however, the undercount has been highest among black people--at least four times as high as it is for whites--and among inner city residents in general. Government demographers say that there is no reason to believe that the pattern will be different this year--a prediction that angers city officials, who say that undercounts will cost their constituencies dearly.

“It is certainly cause for concern,” said Rep. Thomas Sawyer, an Ohio Democrat who chairs a congressional subcommittee on population and has been a cautious critic of census techniques. “This is potentially the first census in the modern era that missed more people than its predecessor.”

A spokesman for Census Bureau Director Barbara Bryant emphasized the “preliminary” nature of the bureau’s October estimate and said time will tell whether it will be a true measure of the undercount.

“Various other studies, which we will do this year, will tell us what the undercount or overcount was,” said spokesman John Connolly. He said he had no reason to question the accuracy of the earlier estimate.

“It was done by a senior official and it is a serious piece of work,” Connolly said.

The census and the October estimate are based on different methodologies. The census is a head count, relying on mailed questionnaires and house-to-house interviews. Its success depended on the cooperation of a mobile, preoccupied, sometimes hostile public. Only 63% of American households completed and mailed back census questionnaires, compared to 78% in 1980.

Census takers had such a hard time gaining access to households that they had to rely on hearsay information about the occupants from 14% of all homes across the country. In spot checks by Census Bureau officials in several cities, officials found numerous examples of undercounting by census takers.

In one South-Central Los Angeles neighborhood, for example, officials determined that census takers had under-reported the number of occupants in 38% of 5,800 households.

Instead of trying to count the total population of the United States, the population estimate tracked birth, death and immigration records. Estimates are subject to change. The original 1980 estimate was revised upward by 3 million people after census officials factored in illegal aliens.

According to J. Gregory Robinson, a senior official in the Census Bureau’s population division, the error rate of past estimates has been between 1% and 3%. Estimates, however, cannot take the place of old-fashioned head counts, Robinson said, because they cannot show how the population is distributed.

The accuracy issue has turned the 1990 census into a political football, as urban Democrats have accused the Republican Administration of presiding over a faulty census that will undercount the residents of big cities, depriving them of their fair share of federal money and political representation.

“At the Commerce Department, statistical grand larceny has become a way of life,” said New York City Mayor David Dinkins. “And the inevitable undercounting, under-funding and under-representation have become as certain as death and taxes.”

Several cities across the country, including Los Angeles, have sued the federal government to force census takers to adjust census figures and base them on a controversial sampling technique, known as the post-enumeration survey. Many population experts believe that the survey offers the best way of counting people, such as illegal immigrants, poor people and transients who often elude census takers.

A settlement of the lawsuit calls for the post-enumeration survey to be conducted and gives the secretary of Commerce, who presides over the Census Bureau, until July, 1991, to decide whether to adjust the census figures based on those survey results.

Census officials say that work on the survey is not complete. In the meantime, 1990 census figures are being released with the admonition that they may be subject to change.

The discrepancy between the population figures released this week by the Census Bureau and the bureau’s earlier estimate led officials in Los Angeles and New York to renew their plea Thursday for an adjustment of the census.

“The census is so far below the estimate, the only thing that makes sense now is a correction using the PES (post-enumeration survey),’ said Jessica Heinz, the deputy Los Angeles city attorney in charge of monitoring the census.