High-speed Internet access and housing values

ABSTRACT

A hedonic model is estimated that relates house values to high-speed Internet access while controlling for the potential endogeneity of Internet access. Results show that single-family homes with access to a 25 Mbps broadband connection have a price that is about $5,977, or 3%, more than similar homes in neighborhoods with 1 Mbps. The rural premium is lower at $5,099. A cost-benefit exercise on the viability of rural broadband shows that demand will generally not support private investment, but that the revenue gap from upgrading legacy networks could be readily covered by the Universal Service Fund and other public subsidies.

KEYWORDS:

• Broadband
• Internet
• real estate
• rural neighborhoods
• universal service

JEL CLASSIFICATION:

• R30
• R21
• L86

Acknowledgments

Lijun Chen, Richard Green, Dale Hatfield, Yang Hu, Milo Medin, Jim Prieger, David Reed, Gregory Rosston, Patrick Thibeault, Thomas Thibodeau, Bradley Wimmer, and seminar participants at the University of Southern California Cable Workshop 2014, WEA 2016, University of Colorado, and the University of Papua New Guinea provided helpful comments and contributions. We are grateful to Matt Drobick and RealtyTrac for assistance with the real estate data, and to the Fiber to the Home Council and Google for providing research support for an earlier version of this paper. Hannah Amundsen provided excellent research assistance. Any opinions expressed in this paper are those of the authors.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Each CBG is comprised of one or more census blocks and a typical CBG has about 610 housing units and about 600 to 3,000 persons. The decision to upgrade Internet service neighborhood-by-neighborhood is common industry practice. For example, CenturyLink deployed high-speed Internet to selected neighborhoods within Denver, while Google defined ‘fiberhoods’ as consisting of 250 to 1,500 households and deployed their fiber service when the demand threshold reached five to 25% of these households (Molnar and Savage 2017).

² Upload speed is becoming more desirable as it is valuable for social networking, cloud services, operating web servers, and for working remotely from the home. However, upload speeds and download speeds are strongly correlated in most Internet service plans and collinearity makes it difficult to isolate their separate effects.

³ See, for example, https://www.geekwire.com/2014/centurylink-gigabit/, https://corporate.comcast.com/news-information/news-feed/comcast-begins-rollout-of-residential-2-gig-service-in-greater-chicago-region, and https://corporate.comcast.com/news-information/news-feed/comcast-to-introduce-worlds-first-docsis-3-1-powered-gigabit-internet-service-in-atlanta-chicago-detroit-miami-and-nashville.

⁴ Household income and educational attainment are typically collinear in hedonic housing models. We also estimated the model with median household income in the CBG (INCOME), instead of EDUCATION, and the results, not reported, are almost the same as those reported in Table 4.

⁵ Rosston, Savage, and Waldman (2010) estimate high-speed Internet valuations for low-income households (annual income less than $25,000) that are about 15% lower than high-income households (annual income of $75,000 or more).

⁶ The signs for most of the significant coefficients in the first-stage have plausible interpretations. ISPs provide higher speeds in neighborhoods with more houses, larger houses (perhaps reflecting Greenfield developments), and with lower crime rates. The positive coefficients on the ISP dummy variables indicate a quality-competition effect in Internet markets. ISPs provide higher speeds in neighborhoods with lower property taxes. This seems counter-intuitive, however, low-tax (‘low-income’) neighborhoods also tend to have overhead lines and cables whereas higher income neighborhoods may have more buried lines and cables, which is more expensive to upgrade.

⁷ We calculate $9,484 from the weighted-average cost per household of building two alternative six Mbps DSL networks (each with different average local loop lengths), and a 50 Mbps cable network. The cost of a six Mbps DSL network is a reasonable proxy for 25 Mbps because the greater portion of costs is for outside plant, which is largely the same regardless of capacity. Higher capacity DSLAMs are needed to provide 25 Mbps, but recent technology change has seen their prices fall to levels comparable to lower capacity DSLAMs from a few years ago.

⁸ When assuming a 30% cost discount, the revenue gap increases to $1,540 per home, and the total subsidy for providing broadband to the seven million unserved rural households is about $10.78 billion. When assuming the 15% demand valuation discount estimated by Rosston, Savage, and Waldman (2010), the revenue gap increases to $1,356 per home, and the total broadband subsidy is about $9.49 billion.

⁹ See http://www.usac.org/default.aspx.