Impacts of the USDA Community Connect broadband program on broadband speeds in rural areas

Abstract

This study investigates the impacts of 23 USDA Community Connect grant projects approved in fiscal years 2017–2019 on measured broadband download and upload speeds during 2019–2022 using program data, Ookla Speedtest® data, and other data. We combine matching and difference-in-difference regressions to estimate causal impacts, using ordinary least squres and two alternative estimators - the Heckman model and extremal quantile regression - to address potential sample selection bias. We find that these projects had a positive impact on upload speed but not download speed, increasing upload speeds in project areas in the preferred extremal quantile regression model using the Mahalanobis matched sample by an average of about 28 percent across the study periods. Our qualitative findings are robust to variations in the study sample, regression model, inclusion of high-dimensional fixed effects or random intercepts and coefficients, and other variations. We find larger impacts on upload speed of projects that supported fiber-to-the-household than other projects and larger impacts in some geographic regions. Based on the preferred model results, we estimate that the program cost of increasing measured upload speed through these Community Connect projects averaged about $1300 per Mbps per household.