Although the three Census 2000 dress rehearsal sites testing census procedures are showing impressive response rates, experts remain skeptical about the outcome.
The Census Bureau began mailing and hand delivering forms for this “mini-Census” to all of the 173,501 households in Sacramento, CA; 280,082 households in the 11 counties around Columbia, SC; and 2,060 households in Wisconsin's Menomonee American Indian Reservation on April 1.
As of April 22, 45 percent of Sacramento's households, almost 45 percent of the households in the South Carolina area and 34.9 percent of the households in the Menomonee Indian Reservation had responded. Residents have until the middle of May to respond.
Preston Jay Waite, assistant director of the Decennial Census, said the bureau is satisfied with the response so far.
“We are hoping to get a 55 percent rate in South Carolina and 50 percent in Sacramento at the end of nonresponse follow-up,” he said. “Previously, in dress rehearsals, we've gotten low 50s and high 40s response rates, so we are certainly on track.”
The rate has impressed some experts — especially since the deadline is several weeks away. But Waite said that since Census Day on April 18, when the bureau used rallies and community events to encourage people to return their forms, the rate of response probably will slow down.
Others are concerned that the dress rehearsal hasn't met or exceeded the 1990 census' 65 percent response rate — even though the bureau ran advertising campaign with Young & Rubicam, New York, at the sites.
Waite said he expects the dress rehearsal to meet or exceed the 65 percent rate.
“The response rate is a very delicate thing and depends on the environment going on in the country at the time we do it,” he said.
Still, some are on the fence.
“It could be on the way to being impressive numbers. If it was all they got, they would have a lot of follow-up to do,” said Ken Hodges, director of demography for Claritas, Ithaca, NY, a consumer marketing information firm that uses Decennial Census data for its products and services. “Clearly, the responses are still coming in — and that leaves room for hope.”
The bureau designed the dress rehearsal questionnaires with larger type and pictures, but officials aren't sure whether that will affect response. In fact, a theme for Census 2000 is “Simple, Less Costly and More Efficient.”
The short form, which was sent to every household in the dress rehearsal and will be sent to every household in the country in the Decennial Census, contains seven questions — the fewest since 1820 and six fewer than the 1990 Census. The long form, which is sent to roughly one-sixth of the population, contains 52 questions, five fewer than in 1990. In both forms, some housing questions have been eliminated.
“The short form will take an average of 10 minutes to complete and is a key element in the Census Bureau's plan to conduct the most efficient, cost-effective census in the nation's history,” said James Holmes, acting director of the Census Bureau.
The 1990 short form took 14 minutes to fill out. This time, the long form will take 38 minutes to fill out as opposed to 43 minutes in 1990. The bureau thinks this will improve the response rate and accuracy of Census 2000. But some are concerned about the questions that aren't being asked.
“The only housing question that will be in the short form is tenure, which means owner or renter status,” said TerriAnne Lowenthal of the Coalition to Preserve Census Data, Washington. “While this is significant for the housing industry, it is also significant for the marketing industry. Marketers need to know if people are [living] in single-family homes or condos to decide who should receive home-oriented catalogs, for example.”
But the simpler census is good news to other companies, many of which use it as the basis for their products and services.
“Anything that boosts the response rates is going to be helpful to the census in general because it will give us better data,” Hodges said. “But it's premature to say if the shorter forms have anything to do with the most recent response rates.”
Jennifer MacLean, vice president of marketing at Metromail Information Services, Lombard, IL, which uses decennial data in its products and services, said the shorter form makes sense — if it gives better response rates.
“A more accurate count is more important to the census than more information on fewer people,” she said. “The higher response attributable to the short form will add to the breadth of Metromail data, while our existing depth will remain strong.”
The controversy doesn't end here. The long form may not even make it to Census 2000. Some Congress members say it isn't needed as the short form is all that is required to collect data for reappointment of the House of Representatives and redistricting.
“In all likelihood, [Congress] will make any decisions about which questions will be included — and if the long form will be printed — before this Congress adjourns in the fall,” Lowenthal said. “Otherwise, they will run the risk of delaying the printing of the questionnaires, which in turn could affect the bureau's ability to start the census on time.”
If the long form is eliminated, “we would get a lot less data, and the quality of some of the products we build in our industry would be diminished,” Hodges said.
Congress is watching the dress rehearsal closely, especially since Sacramento and Menomonee are using statistical sampling and South Carolina is using a nonsampling methodology. Democrats favor sampling, where the bureau tries to count all the people in 90 percent of the households in each census tract — a geographical area consisting of about 1,700 dwellings. Using these figures, it determines the 10 percent of the people who weren't physically counted. The bureau will check for accuracy by conducting a survey of 750,000 households nationwide and making any adjustments.
Democrats say sampling is the only way to count the undercounted, such as inner-city blacks and Hispanic migrant workers. Republicans, on the other hand, favor the nonsampling approach, where the census tries to physically count every person in the country.
The Census Bureau will not know whether sampling worked in the dress rehearsal until enumerators begin following up in late May or early June. At that time, the bureau should know whether it reduced the field work time. A decision on which method will be used in 2000 should come in early 1999.
Sampling also will affect the long-form decision, Hodges said.
“If the critics of sampling prevail — and the 2000 Census has to be completed without sampling methods — it will cost more,” he said, “and some of us are concerned that there might be the temptation to make up for some of that additional cost by eliminating the long form.”