One of the most common questions received here at the atlas is whether there are data for U.S. President tabulated by Congressional District. Many of the inquiries are from visitors whom wish to perform an analysis on how the results of the Electoral College would be different using a district method (such as that currently in use by Maine or Nebraska). The Atlas currently has an incomplete (although growing) set of data for U.S. President compiled by Congressional District for the 2004 general election. The problems facing such a compilation are many.
As an example of the problems encountered, consider the state of Georgia. With thirteen congressional districts to be drawn across 159 counties, the state legislature saw fit (for some reason) to gerrymander the congressional districts such that the boundaries split 33 counties when the decennial redistricting took place following the 2000 census. To compile the results, the data from each of these counties must be obtained at the precinct level and cross-referenced with a key that describes to which district each precinct belongs. However, it turns out that many precincts do not belong to a single district – the CD boundaries actually split precincts. For example, Cobb County has 19 precincts split between districts (either between district 5 and 11 or district 6 and 11). These votes, unfortunately, can not be allocated to a congressional district. Furthermore, all counties in Georgia have a dedicated precinct that includes all absentee ballots cast in the county into a single precinct (this is different from other states, such as New York, where absentee ballots are counted with their individual precincts). In all 33 of the split counties, the absentee ballots can not be allocated to congressional districts. Typically, in 2004, the absentee ballots comprise 10 to 20 percent of the total votes cast in the county. Like absentee ballots, provisional ballots also have their own “precinct” and can not be allocated (although the number of votes in this category is signficantly smaller than absentees or split precincts).
Overall, in Georgia, about 490,000 votes can not be allocated to the correct congressional district (about 15% of total votes cast state-wide). This is a relatively significant figure that may impact the accuracy of the results for President by Congressional District. Its also quite unfortunate for those of us who prefer precicsion in their data.
Presumably such issues would have to be sorted out were a given state to adopt a Nebraska-style allocation system.