Industry-wide, RDD surveys have been plagued by declining response. To address this decline, alternative survey designs have been developed. Many of these designs have been fielded without the benefit of precursor methodological studies. This paper reports on the results of development efforts for a two-phase survey conducted for the Department of Veterans Affairs. The first phase of the survey mailed a short screening instrument that identified Veterans in the household. A second survey was sent to those Veterans identified in the screener. The first part of this paper discusses the results of an experiment that varied the mode of the initial request at the screener. One half of the screener sample was asked to complete the questionnaire on the Web, while the other half was sent a paper questionnaire with no reference to the Web. Web non-respondents were sent a paper questionnaire after one reminder notice.

Data from the completed pilot demonstrate that the mail condition generated significantly higher response rates among both the general public and among Veteran households. The paper will discuss these results, response rates across different types of households, as well as the effects of a paper insert that was included as a second experimental factor. The second part of this paper describes the outcome of having screener respondents choose the mode for the second-phase survey. At the end of the screening interview, eligible respondents were asked whether they would prefer being mailed a paper survey or would prefer the Web option. We will present data on who chose to use the Web vs. mail, the response rate of those that chose the Web vs. mail and the extent to which the Web respondents eventually completed a paper questionnaire (after extensive follow-up).