About the Stanford Election Atlas

Precinct-level data were collected from a wide variety of sources by a team of researchers at Harvard and Stanford led by Stephen Ansolabehere and Jonathan Rodden.  The data were cleaned and matched with precinct-level boundary files at Stanford’s Spatial Social Science Lab, and the interactive map is hosted by ESRI. 

Generous funding was provided by the Alfred P. Sloan Foundation and the Institute for Research in the Social Sciences (IRiSS) at Stanford. 

The data are archived at the Harvard Election Data Archive, where users can download tabular data and boundary files for each state, as well as detailed descriptions of data sources and geo-referencing procedures.  The state-level files contain outcomes of elections for various years and offices.  The online visualization tool only includes 2008 presidential results. 

Please note that this map is not perfect.  For the online map we aimed for the widest possible coverage.  As documented in the data archive, in a small number of instances there were differences in the geography at which county officials reported election results and the geography of the voting tabulation districts on which our boundary files were based.  This was often the product of redrawing precinct boundaries between the 2008 election and the completion of the 2010 census. 

Most of these issues were resolved with the help of county election officials, but in a very small number of localities we had to aggregate votes up to a higher level of geography, such as a ward or municipality, and allocate the votes to precincts according to their voting age population (recorded by the census).  The largest such units were some of the municipalities in St. Louis County, MO. 

In a few states we had to get creative.  Precinct boundaries were unavailable in rural Montana, but we were able to get election results at the level of fragments of Montana State House districts within counties, and we created corresponding boundary files by overlaying precinct, legislative district, and county boundaries.

Rhode Island did not report the geography of its voting tabulation districts, but we were able to geo-code the polling places themselves based on descriptive information about their locations (names of churches, VFW Halls, elementary schools, etc.).  

Finally, due to redrawing of precincts, in Arkansas we were unable to match the 2008 election results with the 2010 voting tabulation districts for which we had boundaries.  Thus we took county-level 2008 presidential votes for Obama and McCain and allocated them to 2010 precincts according to the percentage of countywide votes received by Democratic and Republican candidates to various offices in the 2010 election in each precinct.   

Further details, including sources, are available at the Harvard Election Data Archive.  We are thankful to Dave Bradlee, who provided us with election data for Kentucky, West Virginia, Massachusetts, and Connecticut.

Last modified Thu, 18 Oct, 2012 at 16:19