The Effects of Questionnaire Design Changes
on General Income Amount Nonresponse
in Waves 1 and 2 of the 2004 SIPP Panel
Jeffrey C. Moore
KEY WORDS: dependent interviewing, income amount item nonresponse, interviewer behavior, questionnaire design
ABSTRACT
The “general income” section of the SIPP interview asks about income other than earnings, and other than financial investments – it focuses primarily on participation in and receipt of income from pensions, public assistance, and other transfer-type programs. New procedures were introduced in the 2004 panel wave 2 SIPP questionnaire in an attempt to reduce nonresponse to general income amount items. The new procedures consist of the ability to use dependent follow-up questions of the following sort in the event of an initial nonresponse to an amount question:
“It says here that you received $550 from workers compensation last May. Does that still sound about right for June and July?”
This report summarizes the impact of these new procedures on nonresponse to amount items for selected general income sources in wave 2 of the 2004 SIPP panel, through a comparison with the same estimates from the preceding (2001) SIPP panel. The paper also examines nonresponse to wave 1 amount items for the same income sources. Major findings are as follows:
- The new dependent nonresponse follow-up procedures had a consistent and major positive impact on final nonresponse in 2004 wave 2.
- On the negative side, however, is evidence of improper use of the dependent follow-up procedures by SIPP interviewers, resulting in very high rates of initial nonresponse to the wave 2 amount items in the 2004 panel.
- Despite the fact that income reporting procedures in wave 1 were essentially identical across the two panels, there also appears to have been a consistent decline in wave 1 nonresponse in 2004 compared to 2001. Although the evidence is far from definitive, it appears that a new introduction to the income questions may have helped allay privacy concerns, and thus reduced refusal nonresponse.
CITATION:
