Census-takers in Chicago run into the same problems encountered by a Fuller Brush salesman.
Burly doormen block entry to buildings on the Gold Coast, and tattooed gang members are just as unreceptive when visitors arrive at North Side public-housing projects.
In a test in the city last year, the Census Bureau mailed out questionnaires, then followed them up with personal visits by a census-taker to homes that did not respond. But they were often rebuffed.
"What we've found at this point is that we've had great difficulty getting into buildings at both ends of the economic spectrum," says Bob Marks, the Census Bureau's associate director of decennial census.
The Census Bureau has planned on using polling techniques - known as "sampling" - rather than a door-to-door tally to overcome these and similar problems when it counts the number of Americans in 2000. But those plans have run into their own roadblock: Congress.
Bureau officials say that the census is threatened by provisions in a supplemental spending bill the Senate is considering this week. At issue is a section in the $8.4 billion bill, most of which is aimed at providing disaster relief, that prohibits any fiscal 1997 funds from being used "to plan or otherwise prepare for the use of sampling in taking the 2000 decennial census."
The Census Bureau says that if it can't sample, it can't fulfill Congress's goal of a census more accurate and less costly than the controversial 1990 count. In addition, it says, a ban on sampling would force it to send its "long form" - which now goes to 1 in 6 residents - to every household. The additional work will cost $1 billion, the bureau estimates.
This is more than a simple tiff over tallying techniques. The accuracy of the nation's head count affects the distribution of billions of state and federal dollars for everything from libraries to food stamps. It also is used to draw congressional districts, and therefore can affect which party holds power.
Response rates to census questionnaires vary widely among regions, ethnic groups, income brackets, and housing types, the bureau says. In 1990, it says, it missed almost 2 percent of the population - not a large number, but one that can have a huge effect on the finances of a city like New York, Detroit, or Los Angeles. One estimate is that 4.4 percent of all blacks were missed. A spate of lawsuits ensued, some going to the US Supreme Court.
In the wake of loud complaints from states and cities over the '90 results, Congress appointed a National Academy of Sciences panel to find a way to improve the count. That panel recommended sampling in each county after 90 percent of housing units had responded to census questionnaires. The polling would be used to draw conclusions about the remaining 10 percent.
The bureau has used sampling since 1940 for quality control and improved efficiency, says Martha Farnsworth Riche, Census Bureau director. "This new use would have been an expansion and would have indeed reduced census costs and given us for the first time ever a 100 percent count."
SAMPLING opponents say that the Constitution requires an "actual enumeration," that is, a door-to-door count. In addition, increased use of sampling could result in political manipulation of the statistics, says Paul Trampe, legislative assistant to Rep. Tom Petri (R) of Wisconsin. Representative Petri is sponsor of a bill that would prohibit use of sampling in the allocation of congressional districts.
"To use sampling instead of a hard count is tantamount to throwing out an election because it didn't match the public opinion polls," Mr. Trampe says.
Members of the Senate Appropriations Committee are concerned that sampling will undercount rural areas, says committee spokesman John Raffetto. While the bureau claims its costs will be higher if it can't use sampling, "A lot of members feel that the skepticism that's going to be raised by states undercounted or overcounted by the use of sampling is going to increase the cost to the bureau in resulting litigation," he says.
But Ms. Riche says that door-to-door counts are less accurate than a combination of head counts and sampling. The system works best in small towns and suburbs, she says.
In addition, given working spouses and full employment, it's getting harder and harder to hire enough census-takers, Riche says. "It used to be a census-taker was a housewife, and she went around at dinnertime and pretty much everybody was home. Now it's hard to predict a time when you'll find everybody at home."
The Senate is expected to vote on the legislation as early as today or tomorrow. It is "basically a provision to prohibit planning," Mr. Raffetto says. It only applies until the end of the fiscal year, Sept. 30, leaving plenty of time to discuss the issue, he says.
But Riche notes Sen. Fred Thompson (R) of Tennessee is holding hearings into the 2000 census.
"It seems to me that's the way a decision as serious as this needs to be reviewed, not really as an aside to a process designed to help people whose lives have been devastated by catastrophic flooding," she says.