This working paper provides two contributions by Barry Bye and Sal Gallicchio of the Social Security Administration related to the estimation of variances from the SIPP public-use files. The 1984 public-use data files of the Survey of Income and Program Participation provide pseudo stratum and pseudo primary sampling unit codes that permit direct estimates of sampling errors. The first note is a reprint of an October 1988 Social Security Bulletin article describing a methodology for calculating sampling errors directly from the SIPP public-use file. The authors applied this method to the calculation of variances for persons participating in programs administered by the Social Security Administration, and empirically show an apparent sensitivity of generalized variances (as found in the SIPP Users' Guide and Technical Documentation) to curve fitting procedures.

The second note in this working paper reports the results of comparisons of direct variance estimates from the public-use file with variance estimates based on the original sample design (computed by Census Bureau staff). The authors conclude that the variance estimates are very much alike, suggesting some validity for the direct variance estimates using the pseudo design codes.