Bush faces a different fight over counting

Census Bureau today issues state-by-state count. Should its data be adjusted?

Early next year, George W. Bush will make a potentially historic decision about Census 2000.

The new president must decide whether to permit the United States Census Bureau for the first time to "statistically adjust" the numbers gathered in the nationwide door-to-door count.

If the numbers are adjusted, it could add millions of blacks and Hispanics to the "official" population of the United States. It could also knock millions of other people off the census rolls - people who were apparently counted twice. Most of the people removed would likely come from white households.

Either way, the Bush White House can expect criticism.

Today, the Census Bureau will release the raw, or unadjusted, numbers from the 2000 census for each of the 50 states. The US Supreme Court has ruled that the government must use these actual numbers - not the adjusted ones - to apportion seats in the House of Representatives among the various states.

However, the court left the way open to used adjusted numbers to redraw the congressional district lines inside the states. This could give minorities greater representation in Congress.

Adjusted numbers could also be used to divide up federal aid monies. Immigrant-rich states like California, Texas, and Florida could get billions of additional aid dollars over the next decade if adjusted numbers are made official.

Kenneth Prewitt, director of the Census Bureau, notes that in 1990, the estimated nationwide "net undercount" was 4 million people, or 1.6 percent. That includes an estimated 8 million people who were missed (mostly minorities) and 4 million people who were apparently counted twice (mostly whites).

Dr. Prewitt emphasizes that every census since the first one in 1790 has been merely an estimate. It is literally impossible to count every man, woman, and child, even though thousands of Census Bureau workers this year tried to do just that.

In lieu of perfection, experts have sought ways to improve the door-to-door count by statistical sampling.

Here's how it works.

For this year's regular census, America was divided into 11 million blocks. Some of these were ordinary city blocks, bounded on each side by streets. Others were country blocks that might have rural roads on three sides and a river on another.

The Census Bureau surveyed all 11 million blocks for the regular count. But for the statistical survey, Census workers examined a smaller sample - 11,800 blocks containing 314,000 housing units - randomly chosen by computer.

In July, after the regular census was finished, 6,000 workers returned to the 11,800 blocks selected for the statistical follow-up. Without using any of the data from the first census, they searched out every person in those blocks to find out who was living there on Census Day, April 1.

When the follow-up interviews were complete, census takers had two sets of data for the 11,800 blocks: one from the original door-to-door census, and one from the subsequent follow-up survey.

At that point, the data from the first and second surveys were matched name by name in each of the 11,800 blocks.

Any name that matched in the first and second count was considered a correct entry.

Any name that was found in the second survey, but missed in the first survey, was considered to be an "undercount." Previous studies have shown many of these names are of blacks, Hispanics, and American Indians.

Any name that was found in the first door-to-door survey, but not found by the follow-up, is considered a potential "overcount." To double-check these names, workers returned again to the field - either to track down the missing person, or to determine that the person was incorrectly included in the first census.

Now here's where it all becomes controversial.

The Census Bureau breaks the information from the second survey into 448 demographic units. For example, one demographic group would be Hispanic females, 18 to 29 years old, living in rented housing in rural areas of the West, and who had returned a high percentage of their census forms in the mail.

Having determined in the statistical survey that such women were undercounted by 10 percent in the sample blocks, the bureau would then adjust the numbers for all similar Hispanic women in similar rural blocks in the entire nation.

Comparable adjustments - up or down - are made for each of the 448 groups.

In the Northeast, for example, it might be found that white homeowners, on average, reported too many people living in their households. This could happen because a son or daughter is away at college, and gets counted both at home and at school. Families with two homes also are sometimes double-counted.

In 1990, the estimated overcount for non-Hispanic white homeowners in large urban areas in the Northeast was 2.13 percent. If the statistical method is accepted for the 2000 census, the final official numbers for groups like these could be reduced.

At the Census Bureau, Director Prewitt says he's still operating on orders from the Clinton White House that leave the final decision on using the statistical adjustments up to the agency.

He says: "There's just a lot of exaggerated language and heat about something [statistical adjustment] which, from our point of view, is pretty standard operating practice. If we look at it and we think it doesn't improve the census, we don't use it. And if we think it does improve it, we do."

In a few weeks, Americans will see if George W. Bush agrees.

(c) Copyright 2000. The Christian Science Publishing Society