Technical Difficulties

By Published on .

Call it C2K. There's yet another data glitch lurking beneath America's radar screen. This time, it's one that involves the results of Census 2000. Like the Y2K bug that was the subject of Americans' collective angst two years ago, C2K turns on a single technical anomaly: The race categories used in the latest Census are not comparable to the race categories used in previous decades. On its face, that problem might seem inconsequential. But demographers who first identified the problem say that it may require complex statistical gymnastics on the part of marketers. In fact, those experts warn, if marketers haven't gotten around to worrying about C2K just yet, they will someday soon: “The race issue is by far the biggest problem with Census 2000,� says Martin Holdrich, senior economist at Woods and Poole Economics Inc., in Washington, D.C.

At the heart of the C2K phenomenon is a change in the 2000 Census form that allowed individuals to identify themselves as belonging to more than one racial group. In the past, the Census Bureau has asked people to classify their race as being one and only one category. You could either be black or white or Asian or American Indian, for example. In the 2000 Census, however, you could be black and white and American Indian and Asian, or many combinations thereof. As a result, there are now 57 different race categories — instead of the four single categories that the data world has relied on for decades.

The changeover presents a multitude of problems for marketers. One is a decreased capacity to make historical comparisons. Without being able to make an accurate comparison with the past, a corporation could completely misinterpret the size of its entire market. For instance, a company might think from the data that there is a shrinking population of Asians, when in fact there has been the same number or even a growing number of Asians who now identify themselves as multiracial.

Who will develop a solution for C2K? Private research firms like Claritas recently began formulating their own proprietary “bridge� for making the Census 2000 numbers compatible with existing data. Meanwhile, demographers in academia, the federal government, and private companies have been working on substitute statistical models for some time now. One “patch� will come courtesy of the Census Bureau itself. This summer, the Bureau will survey 50,000 households. First, respondents will receive a mail survey that will be identical to the race question on Census 2000 — people will be allowed to select multiple races. Then, a phone or in-person follow-up with the exact same respondent will ask that person to select just one race.

Another model comes from James P. Allen and Eugene Turner, demographers at California State University at Northridge. Rather than asking multiracial people to pick one racial group over another, Allen and Turner would “fractionally� assign people who picked two or more races into a single race group. Among multiracial people who are both black and white, for example, the Allen Turner method would determine exactly what percentage of the group primarly identifies as white or black. Then, that percentage would be allocated to the appropriate single race group. For example, if 60 percent of people who say they are both black and white primarily identify as black, then the Allen Turner method would simply allocate 60 percent of the total number of multiracial blacks and whites to the single-race black category, and 40 percent to the single-race white group. Allen and Turner tested this model by predicting the percentage of multiracial people that Census 2000 would reveal, and discovered that it was extremely accurate for both black/white and Asian/white people.

Private research firms like Claritas say it will be about a year before they roll out their own proprietary data bridge, and academic and government results may take even longer. Controversy will likely slow the process further — after all, the new models place people into racial categories based on a mathematical formula. But until the debates are settled and the data have a bridge, it's user beware, says Ken Hodges of Claritas. “The new race categories are complicated even for those of us who deal with it all the time,� he says. “We expect some of our clients will be confused for a while.�

Most Popular