The purpose of this project is to demonstrate the use of a cadastral-based dasymetric mapping method to map eligible voter participation at the precinct level. Voter participation can be measured in several ways: 1) the number of votes cast in a given area; 2) the number of votes cast as a proportion of registered voters in a given area; and 3) the number of votes cast as a proportion of all eligible individuals in a given area who could, in principle, vote (i.e. US citizens who are 18 years of age or older). While maps of the first two are common, the last is not. This is due in part to data availability. The numbers of votes cast in a given election and the numbers of registered voters in wards or precincts are data regularly collected and made available by state and local elections officials. By contrast, the numbers of all eligible or potential voters in a given area are often a matter of conjecture or rough estimation. These numbers must be derived by re-aggregating Census populations to voting precincts. However, Census units often do not coincide with precinct boundaries. Traditional methods of map re-aggregation, such as centroid allocation or simple areal weighting, only work well when the census units are evenly spread out, which may not always be the case. In this project we apply a cadastral-based Dasymetric mapping method to estimate eligible voter participation in precincts in Boston, Massachusetts.

The non-coincident boundaries of precinct and tract data make data re-aggregated by centroid allocation or simple areal weighting insufficient for high accuracy population estimation purposes. In this project, to portray the eligible voter participation trend from 2005 to 2012 as accurately as possible, a special dasymetric mapping method is used to overcome this boundary issue.

The potential eligible population consists of people who are 18 years and over, native or foreign born or naturalized U.S. citizens. In order to estimate that population more accurately, parcel property data are used to disaggregate the tract population. In order to generate population estimates of voting precincts, Census Tract population data is first disaggregated to property parcels based on the proportion of living area that each parcel represents within that Census Tract. Each parcel contains information about the amount of living area within the residential units on that parcel. The Census Tract population is allocated to each parcel based on its respective proportion of living area. The parcel level population numbers are then reaggregated to the precinct level.

The re-aggregated eligible voter population data was then used to recalculate voter participation in actual elections between 2005 and 2012. Voter participation in these elections was calculated by dividing the number of ballots cast by the re-aggregated eligible voter population.

One spatial trend that was revealed was that as overall eligible voter participation increased in the city, the differences in participation between precincts decreased (see line chart). In the 2007 and 2011 elections, participation overall reached their lowest points, while the overall differences in participation between precincts were at their most extreme. This pattern held true across all types of elections - municipal, state and federal - and across both Republican and Democratic federal administrations.

Participation rates across time have followed a remarkably similar pattern, regardless of which political party was in power at the federal level. Voter participation rates are highest for all races during a presidential election, and this is also when differences in voter participation between precincts are at their lowest. The year after a presidential election, during a municipal election, overall participation drops to around 25%, revealing significant differences in participation between voting precincts across the city. The following year, during a state or midterm congressional election, participation rises toward 40%, with a concomitant decrease in differences in participation between precincts. Finally, with the return of another presidential election, voter participation reaches a peak of over 50%, and differences amongst precincts are minimal.

Zhenyang Hua | edison.hua@gmail.com
Marcos Luna | mluna@salemstate.edu