Date of Award

5-7-2026

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computing and Information Science

First Advisor

Dr. Hesham Ali

Abstract

This dissertation develops a multi-layer analytical framework to understand electric vehicle (EV) adoption across U.S. counties by integrating (1) temporal correlation networks, (2) predictive modeling, and (3) similarity networks. Existing research often focuses on coarse geographic units, embeds charging infrastructure (CI) within broad multivariable models, or overlooks structural differences among counties. These limitations leave open key questions regarding the directionality of EV–CI relationships, the determinants of county-level adoption, and whether similar counties exhibit similar adoption patterns. The first component examines temporal correlations between counties for EV adoption and CI expansion, as well as Granger-causality relationships under two scenarios: Early Adoption and Late Adoption. The second component applies linear and nonlinear models to quantify how socioeconomic, political, environmental, and infrastructural factors shape county-level EV adoption. Averaged feature importance scores from the top-performing models are then used in the similarity analysis. The third component constructs a similarity network using a Gower distance weighted by these feature weights, combined with a mutual k-nearest neighbors approach. The temporal correlation analysis shows that EV adoption exhibits strong within-state clustering, particularly at finer temporal granularities, whereas CI growth does not, highlighting the complex nature of CI expansion. Granger-causal effects appear more frequently in the Early Adoption scenario; however, when present, the causal responses occur more rapidly in the Late Adoption scenario, likely due to the higher financial and logistical barriers associated with CI deployment relative to EV adoption. Predictive modeling results reveal that nonlinear models substantially outperform linear ones, confirming that EV adoption depends on nonlinear interactions among key determinants. The similarity network further shows that counties with similar weighted feature profiles tend to exhibit similar EV adoption patterns, while also uncovering both global structural trends and local contextual deviations that produce multiple pathways to low adoption. Together, these findings provide a multi-perspective understanding of EV adoption and offer actionable insights for county-level, targeted policy design.

Comments

The author holds the copyright to this work, any reuse or permissions must be obtained from them directly. 

PDF passed Adobe's accessibility checker prior to upload.

Download

Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS