In generating geodemographic models to predict or explain such dependent variables, the approach used is one of observational research rather than experimental -- i.e., secondary data not structured or gathered expressly for the purpose of accounting for the variance in a controlled dependent variable are searched and statistically analyzed in order to isolate efficient predictors.
Fortunately, the scope of secondary data is immense, covering a wealth of topics and showing statistics measuring them for a vast number of small geographic areas such as specific city blocks, neighborhoods, towns, counties or postal delivery zones (ZIP Codes) over the entire United States. For example, a typical "census summary tape file" containing tabulations of the 1990 census "long form" measures the incidence of 3,274 data items for almost one-quarter-million census block-groups as well as hundreds of thousands of other governmentally and statistically bounded areas.
The largest agencies generating secondary data in the federal government include the Census Bureau, Bureau of Economic Analysis, Bureau of Justice Statistics, Bureau of Labor Statistics, Bureau of Transportation Statistics, Economic Research Service of the Department of Agriculture, Energy Information Administration, National Agricultural Statistics Service. National Center for Educational Statistics, and the National Center for Health Statistics. One-half of the federal budget for statistical programs is meted out to these agencies.
about geodemographic data resources...