Given the problems described above, there were two different ways to merge Census and electoral data. One approach would be to match Census VTDs to electoral precincts, producing merged files with the limited set of Census variables available in the PL94-171 datasets; we did not want to start with this approach because we would need to hope that the overlapping pieces of geography were not very important. The second approach, which we adopted, was to aggregate precincts up to the next higher Census geographical unit for which the full suite of demographic variables was available.
The ROAD Project created a number of experimental datasets of the former type, but not for every state. The reason we did not produce more was because (1) VTD definitions are not available for all states or counties and (2) even where fully defined, VTDs are not the same entities as precincts. Hence our VTD-level datasets could not possibly tile the entire country, and the merged data would be relatively suspect even in some cases where VTD definitions exist.
Rather than emphasize what in some cases would be a shaky VTD-level merge, we chose to take the latter approach: to produce a comprehensive national merge of Census and electoral data at a higher level of aggregation (while still providing the electoral data at the precinct level of course). This method had the advantage of including the full range of Census demographic variables. As we shall see, it also made matching of units on either side of the merge much more reliable. Furthermore, this approach also generated a number of byproducts useful for other purposes.
| Copyright © 1997-2004 | [ROAD Home] | Questions? Contact the ROAD webmaster. |