Merging Census and electoral data is not a straightforward task. The central problem is that the Census data does not include a voting precinct geographic unit. Thus, matching Census data to precinct-level voting data for the entire country meant mapping each of the 7 million Census blocks into the 190,000 voting precincts in the U.S.
Fortunately, beginning with the 1990 Census, the Census bureau and the states cooperated to match voting and Census data as best as they could in order to meet legislative redistricting guidelines. As part of PL94-171 (Public Law 94-171), the states (in 1985) suggested to the Census geographic features that would be appropriate for drawing Census block boundaries for the 1990 Census. Then, after the Census delineated the block boundaries and produced maps with the Census blocks to be used in the 1990 Census, the states (in 1989) outlined their current precinct boundaries on these maps and returned them to the Census. Finally, the Census (in 1991) supplied population counts from the 1990 Census aggregated at the voting precinct level using the precinct boundaries supplied by the states.
The result of this process was a precinct-like Census geographical unit, the Voting District (or VTD, which we also refer to as Voter Tabulation District). We quote from the 1990 Census of Population and Housing Technical Documentation:
A voting district (VTD) is any of a variety of types of areas (for example, election districts, precincts, wards, legislative districts) established by State and local governments for purposes of elections. For Census purposes, each State participating in Phase 2 of the 1990 Census Redistricting Data Program outlined the boundaries of VTDs around groups of whole Census blocks on Census maps. The entities identified as VTDs are not necessarily those legally or currently established. Also, to meet the ``whole block'' criterion, a State may have had to adjust VTD boundaries to nearby block boundaries. Therefore, the VTD's shown on the 1990 Census tapes, listings, and maps may not represent the actual VTDs in effect at the time of the Census.
(What is confusing is that in the 1980 census, VTDs were called ``precincts'' even though ``census precincts'' had virtually nothing to do with real electoral precincts. In the 1990 Census, which we rely on, the census units are always called VTDs.)
There are several points to note here. First, census VTDs are not the same as electoral precincts: precinct boundaries were defined locally (usually by county governments), and VTDs were defined by the Census Bureau in cooperation with local governments to closely approximate existing precincts. Precincts are defined for political purposes, and VTDs were created to suit the purposes of the census. VTDs are generally demarcated by natural landmarks, such as roads and rivers, but not so for precincts.
Second, VTD definitions, like all Census area units, were based on the lowest Census geographical unit: the block. Yet the precinct definitions actually used in elections were not based on such a constraint. Thus, in the worst cases, Census blocks are not neatly nested within precincts. As a result, VTDs are not identical to voting precincts and cannot be perfectly reaggregated into precincts. Indeed, on average, VTDs are a bit larger, i.e., more highly aggregated. However, note that in all cases, the perimeter lines for the two aggregation units are never off by more than half a block per road. The vast majority of a precinct, not near the perimeter, is accurately reflected in a VTD.
Third, the Census VTD mapping program was voluntary: certain governments did not participate. Hence some states and some counties lack VTD definitions. VTDs do not completely ``tile'' the nation, i.e., break it into mutually exclusive, exhaustive, and contiguous areal units.
A further difficulty is that Census data at the VTD-level was released in just the PL94-171 set of files. Unfortunately, these files only contain the few population counts useful for redistricting, such as according to voting age and race. The 3,725 other demographic variables of interest to social scientists are not included in the PL94-171 datasets. Hence, in order to use other demographic variables from the Census, one must go to a higher level of aggregation. The next highest level of aggregation that exhaustively ``tiles'' a state is the Minor Civil Division (or MCD), which is also referred to by the Census as summary level 060. There are typically several MCDs within a county, depending on the population of the county. Unfortunately, MCDs only have logically consistent definitions for political and census data for the northeastern states and a few others (NY, NJ, PA, VT, NH, MA, ME, CT, RI, IN, IL, OH, and WI).
| Copyright © 1997-2004 | [ROAD Home] | Questions? Contact the ROAD webmaster. |