This dissertation project seeks to uncover the spatial dimension of information infrastructure at the county level in the contiguous United States and examine its relationship with voter turnout and voting preference in the 2020 U.S. Presidential Election. By leveraging publicly available data from the Federal Communication Committee’s Form 477, the American Community Survey, and Ookla speed test data, an Information Infrastructure Index (I3) was constructed for 3,104 counties, capturing the multi-dimensions of broadband availability, quality, and adoption. Spatial autocorrelation tests confirmed a positive clustering pattern at the county level, prompting the application of the Local Indicators of Spatial Association technique to identify 406 counties as “broadband deserts” and 36 counties as the “fringe of broadband deserts.” These findings underscore the uneven spatial distribution of information infrastructure across the nation, which necessitates targeted interventions by policymakers.
Wielding both global and local spatial models, including the spatial Durbin model and Multiscale Geographically Weighted Regression model, two distinct spatial effects of information infrastructure were identified. Firstly, a global spillover effect on voter turnout was observed, suggesting that information infrastructure from neighboring counties plays a significant role in mobilizing offline political engagement. Notably, the cumulative spillover effect was found to surpass the effect within the county itself. Secondly, a localized non-stationary effect on voting preference during the 2020 U.S. Presidential Election was detected, even after accounting for demographic factors. These analyses provide additional evidence of the partisan effect of information infrastructure, with counties possessing superior information infrastructure tending to lean towards the Democratic Party. While no spillover effect of information infrastructure on Democratic or Republican vote share was identified, the relationship between information infrastructure and voting preference exhibited spatial non-stationarity. Specifically, this partisan effect was more pronounced in certain liberal counties along the East Coast and in the New England area, while the association was weaker in specific conservative areas in the Midwest. Intriguingly, a complete reversal of this pro-Democratic Party effect was observed in regions such as New Mexico, Colorado, and Arizona. These findings contribute novel insights to existing literature by revealing previously unidentified variations in the relationship between information infrastructure and partisan voting preferences in global models.
In sum, this dissertation advances our understanding of the intricate spatial dynamics of information infrastructure and its implications for political behavior. It underscores the theoretical necessity of considering the multidimensionality of ICT within the spatial structure and highlights the importance of adopting both global and local spatial perspectives to yield accurate estimates when analyzing the effects of information infrastructure on political behavior.