Creating SIPP longitudinal analysis files using a relational database management system.

The Survey of Income and Program Participation (SIPP) reflects the growing complexity and size of social and economic microdata files designed to examine a broad range of related policy issues. This paper shows that managing data using relational database management software is a more efficient way of preparing longitudinal analysis files than traditional methods that use statistical packages such as OSIRIS SAS or SPSSX. The authors illustrate how social scientists and policy analysts can achieve large gains in productivity and reduce the large overhead that results from everyone carrying out the same data management operations to create longitudinal analysis files. The facility called SIPP ACCESS located at the University of Wisconsin-Madison maintains the 1984 SIPP in a relational data management system and stores the complete 9 waves of core and topical module data on optical laser disks. Part 1 of this paper identifies the weaknesses of traditional database management strategies for constructing longitudinal analysis files from SIPP. Part 2 illustrates how the relational database management system was used to construct longitudinal analysis files from the 1984 SIPP. Sections A-C replicate the Servais examples. Section D makes use of the longitudinal files that SIPP ACCESS has created to show that only a few lines of English language-like codes are required for constructing a longitudinal analysis file of all persons who ever received welfare and were present in the panel for at least the 1st 8 interviews. Part 3 summarizes the differences between traditional data management using statistical packages and the relational database management system. It concludes with a discussion of how a central data sharing facility like SIPP ACCESS provides the impetus for making major investments in devising efficient cost-effective solutions for improving access to complex data files.