Roberto Gallardo and Brian Whitacre haver released a new research report: An unexpected digital divide? A look at internet speeds and socioeconomic groups. It tracks broadband adoption across most of the US based on speeds and compares to census data. Most of their findings are similar to what we’ve been tracking for the last 20+ years, but there was once surprise…

This study attempted to see if existing disparities among demographic groups in terms of home broadband adoption remained when examining performance or internet quality – measured by average download and upload speed tests from Ookla. The analysis uses over 97% of all census tracts in the continental U.S. and accounts for a wide variety of social and technological factors hypothesized to impact broadband speeds. The results were mostly as anticipated, showing that disparities do exist when looking at internet performance or quality. For example, the share of rural residents has a strong negative impact on realized speed, as does the share of residents under the poverty line. These findings contribute to the digital equity literature and provide timely insights to states and territories going through a digital equity planning process. As additional policies and programs are being designed to address digital inclusion issues, recognition of the disparities that specific demographics face in terms of on-the-ground speed (and not simple availability) is an important part of the conversation. In other words, the assumption that certain covered populations require digital equity interventions may need to be revised when looking at the speeds actually experienced by on-the-ground users.

One surprising result is worth discussing further. The fact that white non-Hispanics were associated with slower average download and upload speeds was not expected. According to the Pew Research Center, as of 2021, 80% of whites subscribed to home internet compared to 71% of blacks and 61% of Hispanics (Pew Research Center, 2021). Similar racial and ethnic discrepancies were found in a recent subscription-focused analysis that incorporated a spatial error model like ours (Zahnd et al., 2022). Yet, the analysis here shows that whites were associated with slower download and upload speeds, controlling for other variables known to affect home internet subscriptions. This finding remains when interacting white non-Hispanic percentage with the percentage of the population that is rural. The estimated impact is not overly dramatic: evaluated at other variable means, the expected download speed would be 171 Mbps for a tract with 10% minorities but 181 Mbps for a tract with 90% minorities. Yet, this is roughly the same impact associated with moving from a tract with 40% of the population over the age of 65 to one with only 10%.

This unexpected finding supports other emerging research. Digital equity surveys in Indiana and Missouri found that whites lag or subscribe at the same rate as racial/ethnic minorities when it comes to paying for home internet for all previous 12 months, after controlling for urban and rural locations (Gallardo, 2023; Spell, 2023). In addition, another study that also used internet speeds across 12 states in the southeastern United States found that the share of whites was associated with slower download and upload speeds, after also controlling for urban and rural locations (Gallardo & Whitacre, 2022). Similarly, a recent study using Ookla data found that majority-Black neighborhoods had higher download speeds during the 2019–2021 period, but slightly slower upload speeds (Rodriguez-Elliott & Vachuska, 2023). Lastly, another study found that while fiber-optic broadband—associated with faster download and upload speeds—increased urban and rural home values in Wisconsin, it also decreased the likelihood of homebuyers being white in urban areas of the state (Wolf & Irwin, 2023). This suggests that minority homeowners place more of a value on faster speed availability—at least in urban Wisconsin. This general hypothesis is supported by our research, but for a much broader geography (i.e. the entire continental U.S.).

It is difficult to explain why slower speeds occur in tracts with more white non-Hispanics. One possible explanation is that this demographic may tend to subscribe to slower speeds when compared to nonwhite consumers, despite potentially having faster connections available. This is not testable with our dataset and remains an area for future research. It may also be the case that some areas that have more internet connectivity issues (e.g., rural) and that conduct more speed tests may also have a higher share of white non-Hispanics. A study conducted in the United Kingdom concluded that rural areas had a higher propensity for speed testing due to network performance issues (Riddlesden & Singleton, 2014). However, the fact that the number of speed tests was associated with higher—not lower—speeds does not support the premise that more speed tests are conducted in areas with unreliable or slower service.