Analytical Comparison of the SIPP and CPS-ASEC Key Longitudinal Estimates

Abstract: The demographic and socioeconomic longitudinal survey SIPP and cross-sectional survey CPS-ASEC have identical survey universes and sampling frames, and are similar in sample design, demographic and socioeconomic contents, questionnaire design profiles, and data collection and processing systems. In contrast, the SIPP collects monthly data but the CPS-ASEC collects annual data. The SIPP sample households are interviewed every four months over three to four years but the CPS-ASEC ones are interviewed once a year for two consecutive years. As a longitudinal survey, the SIPP follows its movers and identifies its survey universe leavers; but, as a cross-section survey, the CPS-ASEC does not. Inspired by similarity between these two surveys, and higher cross-sectional response rates, fresher sample, and establishment stature of the CPS-ASEC, we developed a methodology to construct a quasi-longitudinal dataset from two consecutive CPS-ASEC cross-sectional datasets based on a set of probabilistic models to simulate the survey universe leavers among the sample people in a mix of not-followed movers and universe leavers that could not be definitively identified or separated. We weighted the CPS-ASEC quasi-longitudinal file to reduce the nonresponse bias using the same SIPP longitudinal weighting procedure. As a result of the simulation of survey universe leavers and the longitudinal weighting discussed above, we could use the data of two consecutive CPS-ASEC samples to properly create longitudinal estimates for comparing with those of the SIPP. With addition of identical weighting process to other identical or similar systematic elements mentioned earlier, we could practically attribute the cause of the differences between the SIPP and CPS-ASEC key longitudinal estimates to jus