Missing data are a perennial problem for surveys.  The Census Bureau uses multiple different approaches to deal with missing data, including hot decks, model-based imputation, re-enumeration surveys, and weighting.  However, even with these numerous tools at disposal, missing data persist to the point that underreporting of social safety net receipt is a known and recurring problem for social surveys.  As analysts at the Census Bureau work to redesign the survey, collection, and data editing processes for the Survey of Income and Program Participation (SIPP), this paper explores the comparative efficacy of various imputation processes for these data.