Date of Award
2024
Degree Type
Dissertation
Degree Name
Doctor of Philosophy in Industrial and Systems Engineering
Department
Mechanical, Industrial and Systems Engineering
First Advisor
Gretchen A. Macht
Abstract
Convenience measures for voting continue to evolve in the United States. While most voters are required to cast ballots at predetermined polling locations based on geographically defined boundaries, many states have adopted vote center models that allow voters to cast a ballot at any of the available polling locations in the county either during the early voting period, on election day, or both. A vote center model provides greater flexibility and convenience by allowing voters to choose how, when, and where to cast their ballot. However, the convenience that vote centers provide voters can add a layer of complexity to the election planning process. Throughout this work, different aspects of vote center utilization and implementation are explored to provide insight that is helpful when establishing vote center locations and making resource and capacity planning decisions.
The first work compares voters' demographic characteristics utilizing three different vote center types across the 11-day voting period for the 2020 Presidential Preference Primary and general elections in Los Angeles County, California. This includes comparisons between the different voter center types based on the voter age distribution, percent of provisional ballots, diversity of precincts served, and travel distances. The results indicate that statistically significant differences exist amongst the vote center types for the different demographic characteristics in the days leading up to Election Day itself. The non-conventional vote centers serve as an auxiliary resource to the conventional ones on election rather than their own separate locations and, therefore, have generally similar demographic characteristics. Furthermore, dynamic playback maps are created for the non-conventional vote centers, illustrating the travel behavior of voters to their selected vote center location over the course of each election. This work contributes to the foundational body of knowledge about how voters decide when and where they choose to cast their ballot when afforded options under a vote-center system.
The second manuscript utilizes graph-based methods to analyze network data and community detection statistically. A vote-center network model is established, in which individual vote centers are represented by nodes in the graph and connected based on the voter precincts they mutually serve. Graph-based metrics, including node centrality, assortativity, and clustering, are used to characterize vote center networks for the 2020 Presidential Preference Primary and general elections in Los Angeles County, California. Next, a community detection algorithm is applied to each network graph, and the results are evaluated based on several metrics, including modularity, conductance, and internal edge density. A series of maps are created to provide a geographic representation of the voter center network where the nodes are colored based on their predicted community membership labels, and the line thickness connecting vote center nodes illustrates the strength of their mutual connection. This work provides an interesting way to model in-person voter arrival behavior and statistically test hypotheses about the interconnectedness of vote center locations and the populations they mutually enfranchise. Furthermore, it broadens the intellectual body of knowledge for using graph-based methods to better understand people’s behavior when making decisions independently within an interconnected network.
In the third manuscript, spatial access metrics are utilized to theoretically optimize location selection for establishing vote centers in the Los Angeles County, California, case study. A genetic algorithm that incorporates the spatial access metric into its objective function is used to optimize location selection decisions at the city level in four numerical examples for the cities of Long Beach, Inglewood, Santa Monica, and Downey in Los Angeles County and a fifth, much larger numerical example for Los Angeles County in its entirety. The genetic algorithm efficiently identified the global optimum value for all four city-level numerical examples after searching less than one percent of their total feasible solution spaces, respectively. The global optima were identified through an exhaustive search of each respective solution space. Furthermore, the genetic algorithm showed promise in the much larger county-level numerical example, beating the historical benchmark solution value in all replications with a limited run length of twenty generations.
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.
Recommended Citation
Houghton, James P., "EXECUTING ADVANCED ALGORITHMS TO EXTRAPOLATE VOTER ARRIVAL BEHAVIOR IN U.S. ELECTIONS" (2024). Open Access Dissertations. Paper 1635.
https://digitalcommons.uri.edu/oa_diss/1635