Objective: This paper details the validation of a methodology which spatially allocates Census microdata to census tracts, based on known, aggregate tract population distributions. To protect confidentiality, public-use microdata contain no spatial identifiers other than the code indicating the Public Use Microdata Area (PUMA) in which the individual or household is located. Confirmatory information including the location of microdata households can only be obtained in a Census Research Data Center (CRDC). Due to restrictions in place at CRDCs, a systematic procedure for validating the spatial allocation methodology needs to be implemented prior to accessing CRDC data.

Methods: This study demonstrates and evaluates such an approach, using historical census data for which a 100% count of the full population is available at a fine spatial resolution. The approach described allows for testing of the behavior of a maximum entropy imputation and spatial allocation model under different specifications. The imputation and allocation is performed using a microdata sample of records drawn from the full 1880 Census enumeration and synthetic summary files created from the same source. The results of the allocation are then validated against the actual values from the 100% count of 1880.

Results: The results indicate that the validation procedure provides useful statistics, allowing an in-depth evaluation of the household allocation and identifying optimal configurations for model parameterization. This provides important insights as to how to design a validation procedure at a CRDC for spatial allocations using contemporary census data.