After reading an article suggesting that New York’s subways seeded COVID-19, Salim Furth’s response to that article on this blog, and one or two other pieces, I decided to write a more scholarly piece summarizing the various arguments. The piece is at https://works.bepress.com/lewyn/196/
For those of who you don’t feel like downloading the full paper, here’s a summary:
- Jeffrey Harris of MIT (whose article seeded this controversy) wrote that COVID-19 infections rose most rapidly before subway ridership began to decline; this alone, of course, is not a strong argument because as subway ridership declined, many other crowded places (such as restaurants) were also shutting down. Harris also notes that infections rose more slowly in Manhattan, where ridership declined most rapidly. However, a majority of the city’s jobs are in Manhattan. Thus, Manhattan’s lower subway ridership may have been a reflection not of changed behavior by Manhattan residents, but of the citywide loss of jobs as non-Manhattanites stopped riding the subway to Manhattan jobs. Furthermore, Alon Levy writes that ridership did not decline as rapidly in residential parts of Manhattan (which nevertheless have low infection rates).
Levy also asserts that Harris’s reliance on data from subway entrances is misleading in one technical but important respect. If a Manhattan stops riding the subway to a Manhattan job, this means there are two fewer subway entries for that person. On the other hand, if a Queens resident stops riding the subway to a Manhattan job, this means there is one fewer Queens entry and one fewer Manhattan entry.[ Why does this matter? Suppose that on March 1, there were 100 Manhattan-to-Manhattan commuters and 100 Queens-to-Manhattan commuters, and a week later 30 of each group stop riding the subway. Because there were 90 fewer entries at Manhattan stations (60 from the first group and 30 from the second group), one might think Manhattan subway ridership declined by 90 percent, when in fact it declined by only 30 percent.
2. Harris also relies on the pattern of infections by zip code- and in particular, infections in zip codes along subway lines, because any given rider of a subway line can be infected not only by residents of their own neighborhood, but also by riders who enter at other subway stations on the rider’s route (which perhaps explains why neighborhoods at the end of subway lines tend to have high infection levels). He finds that some subway lines had more drastic declines in ridership than other subway lines- and that the subway lines with more dramatic declines in March ridership also had lower infection rates as of early April. I’m not sure whether the other commentators fully address this point, but maybe I’m missing something.
3. Harris relies on unusually high infection rates among subway workers. Levy responds that subway workers and subway riders do not experience the same risks- subway workers had risks that subway riders did not experience (such as picking up possibly-contaminated rubbish without masks) while conversely, not all subway workers ride packed rush-hour trains.