U.S. ISPs Do Not "Discriminate" in Supply of Internet Access, Based on Census Bureau Data

An analysis of U.S. internet access speed conducted by the Phoenix Center for Advanced Legal and Economic Public Policy Studies does not find evidence of invidious discrimination based on race, in the supply of internet access services. 

The U.S.  Infrastructure Investment and Jobs Act of 2021 says “it shall be the policy of the United States, insofar as ‘technically and economically feasible,’that subscribers ‘within the service area of a provider’ should benefit from the ‘equal opportunity to subscribe to an offered service that provides comparable speeds, capacities, latency, and other quality of service metrics’ at ‘comparable terms and conditions.’” 

The intent is to prohibit “redlining,” where some households in some areas are specifically denied access to some product generally available in a town or city. 

“Digital discrimination, as we define it, is present when differences in some relevant outcome exists across communities when the profitability of serving the communities is equal,” says George Ford, Phoenix Center chief economist. 

The outcome here is not the “take rate,” which is in part a function of end user demand, but “availability” in terms of “a potential customer ability to  buy a product with the same performance and price attributes generally available to potential customers across the market area. 

That is an important distinction. Some consider the existence of any statistical differences in take rates as evidence of discrimination. Differential rates of Tesla ownership, for example, might not be the result of discrimination by suppliers, but rather buying preferences on the part of consumers. 

So “discrimination” implies more than mere differences among protected classes, says Ford.  “Discrimination requires differences in outcomes caused solely by membership in a protected class, other things constant.”

The concept is similar to “redlining,” a once-practiced policy by banks to refuse loans to potential customers in some areas of a city for non-economic reasons (race, generally). 

“Redlining is present when people with equal economic characteristics (who thus may be expected to produce the same “profit” to the firm) experience unequal treatment based on non-economic factors such as race,” Ford notes. 

After analyzing data at the census block level, including downstream speeds, Ford concludes the data show “no evidence of meaningful digital discrimination in any of the comparisons.”

That is not to deny that household income does correlate with take rates. The point is that such divergences are not caused by invidious practices on the part of internet service providers. As is well documented, take rates for premium tiers of internet access service also vary by educational attainment, age or household wealth, for example. 

Still, the average maximum download speeds are approximately 1 Gbps for all groups across all comparisons (race and income), says Ford. That noted, Ford reminds readers that the analysis is at a “whole market” level and does not address individual providers in those markets. 

“Discrimination, at least in economics, is a term of art, and implies differential treatment of equally-profitable groups based on membership in some sort of protected class,” says Ford. The study shows no evidence of such invidious treatment.