Observations on the Census Dress Rehearsal and Implications for Census 2000

This hearing featured Chris Mihm of GAO, accompanied by two GAO evaluators, as the only witnesses. GAO testified on its observations of key census-taking activities performed during the dress rehearsal and the possible implications for 2000. Key activities include address list development, outreach and promotion, staffing, data scanning, and field operations (e.g., the field component of sampling). GAO visited the dress rehearsal sites and observed these census-taking operations first-hand. Following is a summary of their findings and testimony.

    Mail response rates – the percentage of households that returned census questionnaires by mail – at all dress rehearsal sites generally met the Bureau’s expectations. However, while the Bureau’s estimates were accurate, rates remain below what is needed for the Census 2000 plan to work.

    The Census Bureau launched a new initiative in preparation for Census 2000 called “LUCA” – Local Update of Census Addresses. However, LUCA did not work as planned so the Bureau had to “reengineer” its process for creating the address file. This reengineered procedure was not ready in time to be tested in the dress rehearsal and will not be tested prior to Census 2000.

    Local outreach and promotion efforts have suffered a number of setbacks in the dress rehearsal. Many jurisdictions cited a lack of guidance from the Bureau and limited local and Bureau resources as reasons for such limited participation.

    It is too soon to assess the impact of paid advertising at the dress rehearsal sites.

    Staffing the census dress rehearsal has not been a problem to date. According to GAO, “The Bureau attributes this to competitive pay rates, aggressive recruitment, and having a surplus of available employees.”

    Data scanners and other technology have encountered a number of problems (e.g., paper scanners jamming due to accumulated paper dust).

    The ICM phase of the census operation — the type of sampling used to adjust for an undercount or overcount — is ongoing and appears to be on schedule. Non-response follow-up, the first type of sampling used, was generally completed on time or ahead of schedule in the dress rehearsal. However, the accuracy of the numbers will not be known until the end of the year.

J. Christopher Mihm  –  Accompanied by:     Victoria Miller  –  Senior Evaluator GAO
                                                                        Bruce Taylor –     Senior Evaluator GAO