Integrated Public Use Microdata Series International census microdata for social and economic research Sampling IPUMS-USA Sampling practices IPUMS Design: Introductory essays: Sample Design (Chapter 2). Sampling Errors (Chapter 3). IPUMS International For private households, the preferred sampling unit is the dwelling. Group quarters are defined as: institutions and other units with 31 or more persons. Three strategies Where records are ordered geographically: 1. Sample every Nth dwelling or household. For group quarters, the preferred sampling unit is every nth family with associated dwelling characteristics. This systematic sample is equivalent to an equal-probability sample with full geographic stratification. This maximizes precision without the need for a complex sampling scheme. Using the dwelling as the sampling unit reduces the individual-level precision due to clustering. Geographic stratification helps make up some of the loss. If not ordered geographically: 2. Sample randomly 3. Sample by some dwelling level characteristic: For example, the United States Census Bureau uses the following scheme: Race of Householder:   White, Asian & Pacific Islander or Other; Black, American Indian, Eskimo or Aleut Origin of Householder:   Spanish, Non-Spanish Tenure of Housing Unit:    Owner, Renter Family type:    Family with Own Children Under 18    Family without Own Children Under 18    Other Household (non-family) For Group quarters only:    Inmate, Other GQ persons All the combinations of these characteristics are used to define strata. In effect, sampling is done separately for each stratum within each geographic area.