A Method to Audit the Assignment of Registered Voters to Districts and Precincts
The MIT Election Data and Science Lab helps highlight new research and interesting ideas in election science, and is a proud co-sponsor of the Election Sciences, Reform, & Administration Conference (ESRA).
Brian Amos and Michael P. McDonald recently presented a paper at the 2018 ESRA conference entitled, “A Method to Audit the Assignment of Registered Voters to Districts and Precincts.” Here, they summarize their analysis from that paper.
Nearly all United States legislators, at all levels of government, are elected to office by the voters residing in a defined geographic area known as a district. Periodically, district boundaries are redrawn through a political process known as redistricting.
Election officials are tasked with implementing elections using these district boundaries. They associate registrants’ home addresses with districts, and provide voters with ballots that contain the elections they are eligible to participate in. Two recent notable cases in Georgia and Virginia demonstrate that this assignment process is susceptible to error. In these cases, voters were assigned to incorrect districts and voted in the wrong elections; ultimately, the number of affected voters exceeded the margins of victory in the races.
Because voters are in a de facto way being transferred to new districts by election officials, we refer to the phenomenon as administrative redistricting. For the most part, we do not believe these errors are intentional. The task of assigning thousands of addresses to the many overlapping districts is challenging for election officials, particularly those in large urban jurisdictions.
Election officials use two common methods of managing district assignments. In the first method, election officials maintain a master address database that lists every street address range, including identifying odd and even numbered sides of streets, in their jurisdiction. Each street address range is associated with the applicable districts. In the second method, election officials use geocoding applications to pinpoint where a voter registration address is located, and overlay these onto electronic representations of districts.
We develop an automated methodology to audit the assignment of voters to districts, using a methodology similar to the second method, with some enhancements. Most notably, we use multiple geocoding databases, which ensures the greatest geocoding success rate. This methodology is particularly suited to audit the first method, since it provides an external check using an alternative methodology. Our audit methodology can also detect errors in the second method, which are generally due to differences among geocoding databases and subtle mechanics in how geospatial databases are created.
We run our audit on Florida’s voter registration database and find thousands of registered voters assigned to the wrong district. Election officials have responded graciously and professionally to audit reports that we have provided to them, and have worked with us to verify and rectify the errors we detect. We find that urban areas are more likely to suffer from these problems, and by extension, those who are more likely to live in urban areas.
In the course of our communications with election officials, we identify different ways district assignment errors may occur. The first is simple human error, where an election official makes a mistake by assigning the wrong street segments to the wrong district. The second involves geocoding error, where an election administration system’s native geocoder locates an address in the wrong location. The third involves intentional movement of voters. This occurs most often when a district line by law follows a road, the road is physically moved through construction, and affected homes are moved from one district to another. The fourth mistake is an over-reliance on geospatial databases, which occurs when the Census Bureau updates their cartography for errors, but election officials do not, resulting in out-of-sync maps.
To prevent these errors in future elections, we recommend that election officials conduct routinized audits of registered voters’ district assignments, and call for better collection of district boundaries, particularly local boundaries.