The U.S. Census is the mother of all databases. The federal government allocates more than $100 billion in federal funds based on census data. And for marketers, the census “is the benchmark of where the population is, how much money people have, and what the trends are within specific areas,” says Saul Gitlin, director of strategic marketing services at Kang & Lee Advertising, a New York agency that specializes in marketing to Asian-Americans. “If that information is incorrect or inaccurate, it does not allow a company to clearly focus on a potential for its product.”
But some marketers-not to mention government officials-fear that the next nationwide census, to be taken in the year 2000, may not provide a clear sense of the U.S. population. The concerns and controversy hover around sampling (a technique the Census Bureau wants to use to supplement the traditional head count) and the length of the questionnaires.
The taking of the census will begin, explains Census Bureau demographer Arthur Cresce, by mailing to households questionnaires that recipients will fill out an d mail back to the bureau. Then, census-takers will go door-to-door to interview as many nonresponders as possible.
Unfortunately, over the years, the number of forms mailed back to the Census Bureau has declined. In 1980, only 75% of questionnaires were mailed back. In 1990, just 65% were returned. The result: The bureau claims that 1.6% of the population was miscounted or not counted at all.
But many observers-particularly professionals who specialize in ethnic marketing-deem this estimate conservative. “I’ve heard estimates of as much as a 50% undercount, especially in communities such as New York,” says Andrew Morrison, president of Nia Direct, which specializes in marketing to African-Americans and Hispanics.
In hopes of improving accuracy, the Census Bureau plans to add an extra step to its information-gathering process for the 2000 poll. After comparing the results of the questionnaires and the door-to-door interviews, it will determine the number of households it didn’t reach by sampling, a mathematically proven estimate. The 2000 census will be the first to use sampling, but the government has been using it to some degree since 1940; for instance, it uses sampling to calculate the monthly unemployment figures.
“Sampling is going to add dimensions and accuracy to the count,” says Jennifer MacLean, vice president of marketing for information services at database firm Metromail. “You may have segments of the population that may be more or less likely to respond to the mail piece. By taking those nonresponders and getting them into the pool with sampling, it begins to take some of the bias out of the numbers. It makes the statistical community feel the Census Bureau will provide statistical data that is going to be very stable.”
But in February, Speaker of the House Newt Gingrich filed suit against the Census Bureau on behalf of the House to halt the plan, charging that sampling isn’t, in the words of the Constitution, “actual enumeration.” Another challenge, filed by Rep. Bob Barr (R-GA) in Virginia federal court, makes similar charges. Federal court in the District of Columbia is scheduled to hear arguments on Gingrich’s suit on June 11.
Some of sampling’s proponents say that the Republicans’ opposition stems from their concerns that a more accurate count will threaten the House’s 22-seat Republican majority. These proponents contend that minorities, who are overwhelmingly Democrats, make up a disproportionately large percentage of those who weren’t counted. Language barriers, a general suspicion of the government, and fear that the Census Bureau shares data with other government entities, such as the Immigration and Naturalization Service, prevent appreciable numbers of African-Americans, Hispanics, and Asian-Americans from participating.
The sampling issue could remain unresolved up until April 1, 2000-Census Day. This leaves the bureau having to lay the groundwork for two tracks-one including statistical sampling, and one without.
A question of length Another unresolved issue is the fate of the long questionnaire. In 1990, one in six households received a 57-question form; the remainder received a 13-question form. This past April, the Census Bureau submitted to Congress its proposed questions for the census 2000 long and short forms. The proposed short form has only seven questions; the long form, 52.
But powerful members of Congress, including Rep. Harold Rogers (R-KY), who chairs the subcommittee that allocates census funding, oppose the long form altogether. They blame the time it takes to fill out forms for the low rate of mailbacks in the last census. And citing the cost of the census-an estimated $4 billion-critics charge that the collection of this wide range of data is corporate welfare, serving up free data to businesses.
“That simply is not true,” counters TerriAnn Lowenthal, spokeswoman for the Coalition to Preserve Census Data, an ad hoc lobbying group representing businesses and industry groups, including theDirect Marketing Association. “Every question that will be on the long form will include information that is required by federal law to administer federal programs.”
The subjects on the short form include age, sex, race, Hispanic origin, household relationships, and whether the person owns or rents a home. The long form asks about citizenship, ancestry, language spoken at home, plumbing and kitchen facilities, home heating fuel, and vehicles available.
In March, Rep. Constance Morella (R-MD) introduced legislation in favor of retaining these subjects and the long form; the legislation was referred to the Committee on Government Reform and Oversight. The resolution has bipartisan support from 13 cosponsors. Congress is expected to decide on the questionnaires by November, when it is due to award the printing contract for the forms.