Optimized Spatial Census Information Linked Across Time (OSCILAT): Improving the spatial accuracy of 1990, 2000, and 2010 census microdata

Inconsistent geographic boundaries constitute a major challenge for longitudinal analysis of small geographic areas. The boundaries of census tracts and other small areas often change significantly between censuses. To enable direct comparisons over time, multiple data providers offer geographically standardized time series, which employ interpolation techniques to generate multiple years of data for a single year’s reporting areas (e.g., 1970-2010 data for 2010 tracts). These products are nevertheless susceptible to several sources of error, and the magnitude and scope of error are largely unknown. To address these issues, a key resource is the complete census microdata and Master Address File available to approved researchers through the Federal Statistical Research Data Centers (FSRDCs). This information makes it possible to conduct boundary-independent spatio-temporal analysis within an FSRDC and to evaluate errors in public sources of geographically standardized time series. Before unlocking these possibilities, the address information must be mapped to accurate geographic coordinates. We report on a project to use the best available information in FSRDCs to geolocate all of the responses for the 1990, 2000, and 2010 censuses, assigning each record a latitude and longitude as well as 2010 and 2020 block IDs. We document the data, assignment methodology, and high-level results of our project, and we present our plans to share the new data through FSRDCs and to use the data to assess and improve public time-series data maintained by the National Historical Geographic Information System (NHGIS).