I think there are some broader methodological problems too:
The Post adjusted covid-19 rates for cases, deaths and hospitalization over time by combining Centers for Disease Control and Prevention data on cases, hospitalization and vaccinations. The Post used a rolling seven-day average of daily cases, deaths and hospitalization. For vaccination, The Post used the number of people who had received at least one shot as of each date.
For events like covid-19 infection, rates are usually calculated by dividing the number of cases by the number of people in the population. For example, if there are 12 cases among a population of 100 people, the rate would be 12 people per 100. The Post reduced the denominator to exclude most vaccinated people. So if 20 people got vaccinated, that would mean there were 12 cases out of the remaining 80 unvaccinated people, for an adjusted rate of 15 cases per 100 people.
Vaccination is not perfect in preventing infections, however, so The Post did not subtract the entire population of vaccinated people. Data shows vaccines are about 90 percent effective in preventing cases among people who have received the shot. Cases among vaccinated people are called breakthrough cases. To be conservative, The Post estimated that up to 15 percent of the vaccinated population could still be infected.
So, in the example above, instead of removing all 20 vaccinated people, The Post removed 17. That would leave 12 cases among 83 people, for an adjusted rate of 14.5 cases per 100 people.
I think it’s just a straight up mistake to treat people as vaccinated from the day they get their first shot–how could that be more appropriate than a 7- or 14-day time lag? And then they use a single metric for vaccine efficacy, which applies to all vaccines, from the day after the first shot to two weeks after the series is completed. And it appears that they use the same percentage to adjust the lagging indicators, which is pretty hard to defend.
Overall, you’d expect that approach to overstate the unvaccinated case rate: a) in areas where J&J was more common (because it’s less protective than the model assumes), and b) when lots of vaccinations are happening (because you have people in the early stages who are counted as “vaccinated” but aren’t actually protected. And you’d expect it to systematically distort the lagging indicators by counting vaccinations that occurred after the relevant infection.