The Census Bureau recently began a program to field large samples of households, called methods panels, as the vehicle for large-scale experimentation with features of the ACS. The 2006 Methods Panel included 50,000 households and was used to test alternative wording for existing and new questions. A 2007 Methods Panel, which is to include almost 70,000 households, is planned to test not only question wording and questionnaire format, but also strategies to improve mail response (see Box 7-2).
The Census Bureau is to be commended for initiating the ACS methods panels. The program should be continued because of the continuing need for large-scale experimentation on questionnaire format, question wording, instructions for reporting residence, the effects of data collection mode, and other aspects of the ACS data collection. The need for continuing largescale experimentation exists because federal data requirements from the ACS can be expected to evolve over time, as socioeconomic conditions and concerns change. Also, respondent behavior may change in ways that affect
nonrespondents. After 4 weeks, nonrespondents were followed up by CATI; 4 weeks later, remaining nonrespondents were followed up by CAPI. There was no telephone questionnaire assistance or telephone edit follow-up, which could have influenced respondents’ answers. After data collection, a subsample of mail, CATI, and CAPI respondents who furnished a telephone number were followed up by CATI to measure simple response variability and response bias by comparing answers from the first interview (by mail, CATI, or CAPI) and the second CATI interview.
Analysis of results and recommendations for new and revised content for the ACS beginning in 2008—expected in early 2007.
2007 ACS Methods Panel
The 2007 ACS Methods Panel is designed with two tracks:
The first track will address new and improved content, including a new question on field of bachelor’s degree and a modified format for the basic demographic questions (age, sex, race, ethnicity, household relationship). Four different questionnaires will be mailed to a total of 30,000 housing units, with CATI and CAPI follow-up and a CATI content reinterview.
The second track will address ways to increase mail response and thereby contain costs. One strategy for testing is to make another mailing to nonrespondents for which a telephone number is lacking (three different mailing pieces will be sent to 6,000 housing units each). Another strategy for Puerto Rico and targeted areas of the United States with the lowest levels of mail response is to include a motivational piece in the questionnaire package. Two different mailing pieces will be sent to 10,000 housing units each in the targeted areas in the United States.