This is just a census test

American Demographics, Feb, 1996 by Diane Crispell

Next month, the Census Bureau will conduct a test census to find out how Americans might respond to a newly designed census package.

The U.S. Census Bureau is under immense pressure to conduct an improved census in 2000 for less money. Toward that end, it's gearing up to try some new things that may increase the initial mailback response rate from its less-than-stellar level of 65 percent in 1990. A big part of the bureau's hopes rest with newly designed user-friendly questionnaires.

The 2000 Census Test, formerly known as the National Content Survey, will be conducted on March 2. It's not the first test the bureau has conducted for the upcoming headcount. In fact, seven census tests were conducted between 1992 and 1995 to examine issues ranging from machine readability to the effect of questionnaire length on response rate. But March 2 is the big one. The bureau will send short-form questionnaires--now known as "simple" forms--to about 42,000 households, and long-form questionnaires--now called "sample" forms--to about 52,500 households. The test results will help the bureau make decisions about specific question wording, overall format, question sequence, and form design.

Sample households will receive one of seven simple forms or one of six sample forms. Two are controls that duplicate the 1990 census short-form and long-form questionnaires. The Census Bureau designed three of the six noncontrol simple forms. The other. three are experimental. forms; New York City-based design firm Two Twelve Associates designed two, and consultant Dr. Don Dillman of Washington Pullman. Washington, designed all five the noncontrol sample forms, although a commercial contractor evaluated and simplified the wording on one of these.

The commercially designed sample forms are a visually striking departure for the Census Bureau. The Dillman form is green. Two Twelve chose yellow because it is rarely used by the commercial direct-mail firms that now compete with the census for Americans' attention. Both colors meet the bureau's image-processing requirements. The forms also use graphic design to identify the package as federal documents. The envelopes carry through the color/graphic theme as part of an integral package, just like commercial direct mail.

All three experimental forms are simpler and clearer than those sent in 1990. For example, they have either shortened or completely eliminated the instructions about who to include or not to include in a household count of people. Each has a clearly designated block of space to list individual information on up to five people per household, compared with the sometimes confusing matrix format used in past censuses.

The Census Bureau doesn't intend to ignore large households, however. On all three experimental forms, respondents are instructed to provide a total count of household members, and two forms have space to list the names of those for whom individual data are not provided. The bureau will then send enumerators to follow up with this small group, which is probably just as well, since they tend to be poor respondents. The 3 percent of U.S. households with more than five people in 1990 tended to have lower levels of education and limited English language skills, both of which contributed to a lower-than-average response rate.

It's unclear how the experimental forms will affect the response rate and quality of information for large households. Of course, this is one of the reasons why the bureau is testing them. One advantage of the five-person limit is that it shortens the forms. Questionnaire length is one of the few previously tested features that has a negative effect on response rate; the longer the form, the less likely people are to fill it out.

Previously tested respondent-friendly questionnaire designs increased response rates by a significant but small 3 percent, and the bureau hopes that the experimental forms being tested next month do even better. Other factors have tested more strongly, including an advance notice letter that tells people they will be receiving a census form (6 percent increase), a reminder postcard sent a few days after census forms (8 percent increase), a replacement questionnaire sent about three weeks later to nonrespondents (10 percent), and an envelope message stating that responding to the census is required by law (10 percent). The March content test will incorporate all of these features. It will also include some that have not proven to improve response rates but are still considered important, such as a motivational message that explains in plain language how communities benefit when residents participate in a census.

One of the many things the Census Bureau hopes to accomplish with jazzed-up census forms is to improve a somewhat tarnished public image. But its main goal is to get more Americans to answer the census in a timely manner to save money on follow-up contacts.

For more information about the 2000 Census Test, contact Patricia Berman, Decennial Management Division, U.S. Census Bureau; telephone (301) 457-3960.