In the general discussion on priorities for data collection, one topic that emerged had to do with the frequency with which surveys should be carried out. Given the impracticalities of having high-frequency surveys in the field on a continuous basis, several participants suggested that it might be useful to target a subsample of a major survey (such as the HRS) for more frequent interviews, perhaps identifying vulnerable populations that were experiencing or were at risk of experiencing certain events and then following up with them at short notice for more detailed interviews. The success of such an effort would hinge on the ability to establish protocols and procedures for going into the field at precisely the right time.

One participant suggested that the experience of the ALP, which came about as part of the project on Internet interviewing and the HRS, may be instructive. One of the purposes of the project was to make the HRS instrument suitable for Internet interviewing, and all HRS modules are now being administered to ALP respondents. Roughly two surveys are administered per month, and they are often fairly short (around 20 minutes) and do not seem to be onerous; response rates are around 80 percent. The ambition is to expand the size of the sample (currently around 2,500), which will make it easier to experiment and do other things with the survey.

The importance of communication and collaboration among different surveys, as well as the need for flexibility and innovation (without necessarily having such experimentation being tied to particular projects), was also noted. For example, something like the MINYVan survey could be used as a test bed for questions that ultimately end up in a special module of the HRS. Or there could be an ongoing test base study that receives input from multiple contributors, encourages collaboration and experimentation, and contains ample room for error.

One participant commented that time-use data are fairly sensitive measures of what happens during a recession (whether people spend more time watching television, looking for a job, etc.) and could be useful if they were reported frequently enough. Participants also discussed the potential for merging data sets with proprietary information such as mortgage and credit records. Informed consent was mentioned as a serious issue, as people could provide approval but then forget that they did so; very general consents also might not hold up for more specific purposes. Scandal in one data set can very easily spread to others, causing response rates to plummet. It was suggested that these issues might best be approached through pilot surveys that are separate and distinct from other ongoing surveys. Two other issues that were raised relating to merging data sets with proprietary information were the operational risks of data getting lost or misplaced during transfers and the problematic nature of geography as a potential identifier.

One participant brought attention to two data systems that were not mentioned during the workshop sessions. First, the 2007 wave of the Survey of Consumer Finances has been made into a panel, with reinterview information collected in 2009; the panel data will be made publicly available in early 2011. Second, the 2008 panel of SIPP began in fall 2008 to collect detailed information from the previous 4 months. Because of supplemental funding from SSA, SIPP administered topical modules on pension coverage and retirement accounts in summer 2009 and assets and liabilities in fall 2009.