Global stay-at-home orders in response to COVID-19 have shifted even more of people’s daily lives onto the internet, and providers have struggled to keep pace. Existing research on broadband internet usage and demographic data is sparse, but analyzing those relationships in the context of a pandemic is unprecedented. Jonathan Williams, an associate professor in the Department of Economics at the University of North Carolina at Chapel Hill, recognizes this.
Throughout March, classrooms and workplaces rapidly adapted to online platforms like Zoom. How providers have handled this increase and how different demographics have adapted to these education and work transitions, is William’s focus.
Williams runs the Center for Regulatory and Industrial Studies (CRIS), which maintains working relationships with industries the center’s faculty are interested in researching, like airline and cable, in order to obtain useful data and insight in exchange for advice and analysis. Williams has watched the two industries go in opposite directions. “Half the airline staff I work with now has been furloughed,” he says. “Whereas the telecom guys, they can’t hire people fast enough.”
Industries are looking to researchers like Williams for answers to pandemic-related policy and regulation issues. Williams and his colleagues at CRIS recently received novel data on household’s broadband usage and demographics from providers eager for analyses.
“Any insights they can get are helpful for them,” he says. “And that’s why they were so quick to throw this data at us.”
For example, now that the Federal Communications Commission (FCC) has rolled out a program for larger operators to guarantee broadband to families regardless of ability to pay, they are interested in how many families are homeschooling and how families’ usage differs among demographics. This may demonstrate how the pandemic may exacerbate education inequality between income levels and racial groups.
“There’s just a ton of questions that they care about from a policy perspective,” Williams says. “And I think the FCC is spending a lot of dollars to try to address that.”
Williams’ analyses show that, from mid-February to early April, daily internet usage per household increased by 31 percent. Beyond the general increase in volume, big shifts in the composition of internet traffic has posed challenges to providers. Upstream traffic associated with video conferencing software like Skype and Zoom increased by over 46 percent. While fiber-to-the-premise networks — which use fiber optic cables — are less likely to have capacity issues, DSL and cable networks can struggle to adapt to these rapid shifts.
The cable industry’s networks of cables, for example, are set up asymmetrically, meaning the majority of normal internet traffic goes in the direction of the user. Most users are downloading rather than uploading. They consume data, or have data “pushed” to them, through video downloads and webpages. Cable operators provision their networks for this asymmetrical demand. Speeds may be 750 megabits per second for downloads and 25 megabits per second for uploads. But lately, demand is not so unbalanced.
“When you think about something like Zoom, Skype, or any of these real-time communications applications, they’re very symmetric,” Williams says. “You’re pushing as much data as you’re consuming. You’re getting video from other people, but you’re pushing video to them as well.”
The load of symmetric demand by telehealth, conferencing, and online schooling is not the only issue. Internet providers that have tailored networks to handle peak demand during the evening are now facing usage that surges early and plateaus through the entire workday.
“For them, it’s all happened in a matter of weeks,” Williams says. “And a lot of these networks require longer-term adjustments to be able to adapt to those types of shocks.”
The data Williams received are essentially the hourly usage of 80,000 households, including what users were doing during each hour such as real-time communication on Zoom or streaming on Netflix. The data includes the introduction of stay-at-home orders, so Williams can see how the load on the network changed over time.
Williams is especially interested in the comparison of demographic data. When signing up for services like broadband, people typically undergo a credit check. This means that usage data can be matched up with household details made available in corresponding credit information like number of children, head of household, gender, income, and race.
Right now, Williams is working with his co-authors, second-year PhD student Haoran Zhang and University of Pennsylvania economics professor Aviv Nevo, to parse through the data and see how usage patterns between different demographics shifted since the pandemic. They will analyze the data to understand how the online shifts in work and education have impacted, and perhaps worsened, income and education inequality.
Despite the difficult circumstances of COVID-19, Williams says that the data may reveal silver linings. Companies may discover that some employees are more efficient when working from home and educators may find that certain students learn more on their own. Williams also hopes that this increased demand on the industry can accelerate investments in upgrading broadband networks.
Jonathan Williams is an associate professor in the Department of Economics within the UNC College of Arts & Sciences.