When Census Geography Doesn’t Work: Using Ancillary Information to Improve the Spatial Interpolation of Demographic Data

Voss, Paul, David D. Long, and Roger B. Hammer
Working paper no. 1999-26


This paper introduces two new spatial interpolation techniques that utilize the network of road segments and the resulting nodes to allocate aggregated demographic characteristics from one type of geographic boundaries (i.e., the geographic hierarchy of the U.S. Census) to another (e.g. watersheds) under conditions of “spatial incongruity.” Spatial incongruity arises when spatially aggregated data are available for one set of geographic areal units but not the areal units of primary interest. Spatial incongruity presents a major obstacle to the integration of social and natural science data and consequently places limitations on interdisciplinary research efforts. In the natural sciences the geographic units of analysis frequently are areas defined by land use, land cover, soil type, watershed boundaries, and a variety of other biophysical and geophysical features. Given that census geography and its concomitant demographic data seldom correspond exactly to these areas, combining the data from different disciplines and disparate units of analysis becomes a crucial function. The road segment length interpolation method presented in this paper improves upon areal weighting, the most common method used to allocate characteristics from one geographic system to another, in limited circumstances while the nodal count method represents a substantial improvement.