Children's Health Insurance Patterns: A Review of the Literature. 1. Overview of the SIPP


The SIPP is a multipanel longitudinal survey of adults in a sample of approximately 20,000 households selected to be representative of the noninstitutionalized resident population of the United States. The research reviewed below is based on data from either the 1990, 1991, or 1992 SIPP panels. These panels followed sampled adults for approximately two-and-a-half years, interviewing them either in person or by telephone every four months.(12) During each SIPP interview (called a wave), household-, family-, and person-level information is collected for each of the previous four months on income, labor force activity, program participation (such as AFDC, Food Stamps, and Medicaid), and health insurance status.

Over the life of a given SIPP panel, the Census Bureau produces a separate data file for each wave. At the conclusion of the panel, the Census Bureau produces a full panel file containing one record for each person who was ever a member of a SIPP household. This file differs in two important ways from what users would create were they to link each of the wave files for the panel. First, an entirely separate set of edit and imputation procedures is used by the Census Bureau when creating the full panel file, including the imputation of missing waves when a single missing wave is bounded on both sides by reported information for that person. Because these data processing procedures take advantage of the reported longitudinal information in the file, the full panel files are generally believed to have superior data to those contained in the wave files. Second, the full panel files contain weights that are not available on any of the wave files. The Census Bureau generally creates two types of weights for its longitudinal files: (1) calendar year weights for all persons present throughout a given calendar year of the panel, and (2) panel weights for persons present throughout the entire panel. All other persons do not receive weights and, thus, are not counted in longitudinal estimates.(13)

Like all longitudinal surveys, not all respondents remain in the survey for all the interviews. This is known as attrition and may bias estimates using the SIPP to the extent that those who attrite are systematically different from those that do not. Research on this topic has indicated that for many estimates, there is no detectable bias that can be attributed to attrition. Still, more research needs to be conducted before firm conclusions can be drawn on the effects of attrition in the SIPP.