A woman works from home, in Vertou, outside Nantes, France. On average, US activity was up 16% in April and 23% in May 2020, compared to expected activity based on historical data. Assuming a baseline of 40 hours per week, that translates to an additional six hours and nine hours per week in April and May, respectively. — AFP
How the pandemic change work schedules? We can look to Github data for one answer.
At the onset of the Covid-19 pandemic, Wenqi Shao, a San Francisco data scientist, shifted her working hours in response to her suddenly flexible work-from-home schedule. Shao, who was then working for the shipping technology company Flexport, opted to work later in the evenings, freeing up time in the middle of the day to exercise or run errands.
Shao is one of many developers who shifted their work schedules in response to remote work. On average, developers in San Francisco worked an additional two hours per week outside of the traditional 9am to 6pm weekday schedule – putting in more work during weekends and evening hours.
That’s according to a new working paper by two economics professors from the University of Oregon, which uses data from the popular development website GitHub to examine the pandemic’s effect on activity among GitHub users.
GitHub is the world’s largest provider for software development. It is used by over 73 million people worldwide, including professional software engineers, researchers, students and government agencies (The Chronicle also uses it for its journalism). Its main purpose is version control and collaboration. A group of developers can view, edit and upload different versions of the same code or data file – like if Dropbox and track changes had a baby, as Grant McDermott, one of the researchers, described it.
McDermott and his co-researcher, Benjamin Hansen, gathered data on public GitHub activity in 2020 and compared it with activity prior to the pandemic (2017-2019). Public GitHub activity refers to every event that is recorded on a public repository, such as when a user uploads (“pushes”) a change. Using this data, the researchers compared the share of activity that occurred during weekends or outside of the traditional 9am to 6pm weekday work hours.
They found that, in San Francisco, weekend and out-of-hours work spiked at the onset of the pandemic. The share of all activity that occurred during the weekend increased from a pre-pandemic average of 17% to over 20% in March, while the out-of-hours percentage increased from 32-33% to 35%.
The researchers translated the change in activity to hours worked. Assuming a 40-hour work week, these increases in San Francisco translate to an additional two hours per week during the weekend or outside of the normal 9-to-6 on weekdays.
There are countless reasons for the shift in work schedules, said McDermott. Some people may have experienced increased workloads due to the virus. Others might have needed to work more hours on the weekend because of additional responsibilities at home caused by the pandemic. As a parent, McDermott recalls devoting much of his normal working hours on childcare and not being able to work until later in the day.
In addition to shifting schedules, the researchers found an overall increase in GitHub activity. On average, US activity was up 16% in April and 23% in May, compared to expected activity based on historical data. Assuming a baseline of 40 hours per week, that translates to an additional six hours and nine hours per week in April and May, respectively.
D’Arcy Loeb, a product manager at the San Francisco-based healthcare technology company Mobile Health, told the Chronicle her hours increased to roughly 60 hours per week in April and May. When the pandemic hit, she worked with a team of developers to create a Covid-19 screening app that companies used to track when and if employees can return to work after being exposed to the virus. Having to create the product almost overnight and keeping up with the constantly-changing CDC guidelines led to an astronomical amount of work, she said.
But according to the GitHub data, these elevated activity levels peaked in early May but fell to normal levels in the following weeks. By the end of June, the average activity level was roughly 10% more than normal. This return to pre-pandemic trends coincides with when parts of the economy reopened and some out-of-home leisure activities returned.
By the end of 2020, a “new normal” seems to have emerged, said McDermott. “People appear to be working more during traditional leisure times – but not necessarily more in the aggregate – as schedules adjust to a world where more and more of us are WFH,” he and Hansen wrote in their paper.
It’s worth noting that this data represents a small – and generally privileged – share of the workforce. GitHub is often used by software developers who typically experience high salaries, job security and had work-from-home options available to them throughout the pandemic.
But 2022 may be the year that developers and other remote workers return to offices. Despite the Omicron variant currently sweeping through the Bay Area, some medical experts posit that the virus will largely be managed at some point this year, thanks to widespread vaccination and immunity.
But companies that delayed return-to-office plans because of Omicron have yet to announce new return dates. Some have even announced permanent remote work options. This suggests the shifting work schedule found in McDermott’s research may persist even after the “end” of the pandemic.
For you data nerds who are curious about the research methodology:
Analysing real-time data like GitHub activity is challenging due to the fluidity in the number of users that generate data points. For this study, the researchers needed to control for the increase in GitHub users over time when calculating and comparing activity levels. To deal with this issue, they trained a time-series model on historical activity (2017-2019) and had the model simulate a counterfactual for 2020.
In other words, if historical patterns persisted, what would we have expected 2020 activity levels to be? That’s how the researchers produced the “predicted” numbers, which they then compared to the actual data to understand the difference. – San Francisco Chronicle/Tribune News Service
(Nami Sumida is a San Francisco Chronicle data visualisation developer.)
