Evaluating the Impact of USDA Broadband Subsidy Programs Using Differentiable Gaussian Random Fields

• Activity Number: 152 - Statistical Methods for Data Privacy and Statistical Modeling of Social and Economic Factors
• Type: Contributed
• Date/Time: Monday, August 8, 2022 : 10:30 AM to 12:20 PM
• Sponsor: Social Statistics Section
• Author(s): Joshua Goldstein* and Aritra Halder and Hanna Charankevich and John Pender and Aaron Schroeder and Stephanie S Shipp and Sallie Keller
• Companies: Biocomplexity Institute, University of Virginia and University of Virginia and University of Virginia and Economic Research Service, USDA and University of Virginia and University of Virginia and Biocomplexity Institute, University of Virginia
• Keywords: Bayesian hierarchical model; broadband programs; Gaussian Processes; gradients; spatial regression discontinuity; Wombling

Abstract

The US Department of Agriculture’s Rural Utility Services (RUS) administers grant and loan programs in rural areas to improve infrastructure access and quality of life. We develop statistical methods to study the impacts of RUS broadband programs on residential property values as well as the differential impacts on race and ethnicity. Our approach features a Bayesian hierarchical spatial modeling of property prices at its core, which produces an estimate of the underlying spatial random field for eligible geographic regions. This is followed by algorithmic detection of significant regression discontinuities and gradients along program boundaries within the random field. These estimates reflect the rationale that if a particular program is effective, the increased availability of broadband should affect differences in property prices across the program boundaries. We apply these methods to evaluate impacts across many rural project service areas as part of the Broadband Initiatives Program.