CENSUS 2000: Mailers Voice Views on Sampling
Sampling is a way to determine the number and characteristics of people not reached by more conventional methods, such as door-to-door interviews or questionnaires.
Similar to methods used by public opinion polls, sampling extrapolates the characteristics of large group by surveying a representative sample. Sampling, its endorsers say, includes those historically missed by the government -- predominately the poor and minorities. Recent studies show that only 61 percent of households answered the 1990 census questionnaire.
The business and direct marketing community who use the census in their products or for targeting purposes have mixed feelings about sampling.
"I advocate sampling," said Ken Hodges, director of demography at Claritas, Ithaca, NY, a consumer marketing information firm that uses census data as the basis for many of its demographic products and services. "It's scary. It's risky. It's never been done in a full decennial census before, but my view is that the only thing scarier than a census with sampling is a census without sampling."
Hodges pointed to the importance of sampling for the "follow-up for nonresponse" portion of the 2000 Census. According to the administration's plan, census employees will attempt to collect a questionnaire from every household, but then contact a statistically representative portion of those households that fail to respond.
The results of that sample would be used to arrive at a total population figure, along with the race, income and other characteristics of those who could not be counted directly.
"When people don't send their forms back, the census has to do follow-up that is very labor intensive, very expensive and yields some of the worst quality data you can get," he said. "Sampling would improve this a great deal."
In addition, Hodges said, sampling would arguably give his direct marketing customers better, more high-quality data.
Some mailers say undercounting minorities doesn't directly affect them -- yet.
"Of the direct mailers, the ones to whom this is particularly important, are the political mailers," said Jan Davis, vice president and general manager at the PerfomanceData division of Trans Union Corp., Chicago. "For the rest of us, it is less of an issue. Most of the areas where there is believed to be under-representation in a true census are either very poor neighborhoods or new immigrants -- who are not likely targets for direct mail."
Two dress rehearsals are planned for March involving different collection methods for Census 2000.
One will take place in Sacramento, CA, where the Census Bureau will mail out forms, and depending on how many are returned from a given neighborhood, send head counters to the residences of a sample of those who did not respond. The bureau then will extrapolate the number and characteristics of those who did not return the forms.
The other will take place in a rural area in and around Columbia, SC, where the bureau will use more traditional methods, i.e., it will send out several mailings and then count door to door to find those who did not reply.
No one is sure if sampling will actually be part of Census 2000. While the National Academy of Sciences, the National Research Council, the American Statistical Association and the General Accounting Office all endorse sampling, Republicans and Democrats are squabbling over it. To be validated as a method, agreements must be reached in both the House and Senate.
All parties -- the business community included -- agree, however, that the bureau needs a leader.
"The business community has a major stake in the quality of the 2000 Census," Hodges said, "and given the fact that we are approaching a critical time in the development and execution of that census, the fact that the bureau is without a director is a real cause for concern."