An evaluative study of geo-demographic database systems

The purpose of geo-demographic database systems is to develop a standardized classification which serves to identify homogeneous population groups according to similiar socio-economic and demographic characteristics. The census tract and ZIP code areal unit are normally utilized. Some applications of geo-demographic database systems are for trade area analysis, target marketing, and demographic studies. These systems represent an improvement over more classic urban structure models. They are based on sophisticated factor analytic and clustering techniques using a greater range of variables than classic urban structure models permitted, and are compatible with a variety of geographic scales. The thesis examines how these systems are developed, the methods used to develop these systems and how results are interpreted. A case study of the Knoxville MSA is undertaken at the census tract level. Using a data set developed by the United Way of America, two research hypotheses are tested. Hypothesis one examines the sensitivity of group assignment based on the factor analytic and clustering techniques utilized. The results indicate that group assignments are sensitive to both the type of factor analytic and clustering techniques utilized. Hypothesis two examines the sensitivity of cluster group assignment to the input variables used. To test hypothesis two, I developed an extended data set merging the United Way data with 23 variables selected from previous factorial ecology and quality-of-life studies. The results indicate that the extended data set offers dramatic improvement in the amount of explained variance. Furthermore, cluster group assignment is more easily accomplished with this extended data set. Using standardized factor loadings, cluster means and standard deviations are calculated and the resultant clusters labeled. Geo-demographic database system output based on the census tract areal unit is useful in providing general information about neighborhood composition. The results also indicate that the census tract is too heterogeneous to accurately segment the population into homogeneous groups. In the Knoxville MSA, some significant population groups existed within census tracts whose characteristics differ startlingly from the average tract composition and might, therefore, been overlooked. This aggregation problem may be solved if the data set used to develop the system could be based on forthcoming Census 90 data. Census 90 data will allow one to create smaller, more homogeneous areas than the census tract which should improve overall system usefulness.