Three key steps in the development of procedures for obtaining unbiased longitudinal household estimates of population totals in the Survey of Income and Program Participation are:

1. Select a definition of household continuity.
2. Develop a method for creating longitudinal household files with the definition selected in Step 1.
3. Determine a procedure for obtaining a set of weights for the longitudinal household files created in Step 2 which yield unbiased estimates.

(In this paper adjustments to the unbiased estimates are not discussed. Also, such problems as imperfections in the sampling frame, nonresponse, and response errors, which in general result in truly unbiased estimates being unobtainable in practice, are ignored. It is assumed that the longitudinal universe is a partition of the union of all households in the cross-sectional universe for one or more months during the interval under study, except those cross-sectional households which contain no people who were in the cross-sectional universe at the time of the first interview.)