A few points. One is that Nate Silver (former bookie and now runs Five Thirty Eight) has a pretty good write up
Corona Case Counts are Meaningless. His point is that number is sampled and calculated differently in different places, and is not a very good indicator of what is going on.
Second,
@Astronuc seems to have taken issue with my claims that the Washington flu numbers are underestimated (although they are probably underestimated by the same factor year to year). So here's my comparison between flu and Covid.
1. Europe:
Thanks to
@Stephen Tashi , we have total death statistics from Europe from all causes. Europe has been both harder hit and is in a later stage of the epidemiology than the US.
That uptick in 65+ and Total at the very end is Covid. The downtick in the other age groups is hypothesized by
@russ_watters to be a side effect of lockdowns. You can see the spikes (which look like lumps) in the last three winters. The CDC US death estimates for flu for those years was, starting with 2018-9 and working backward, 34K, 61K (in two phases) and 28K, these are all fairly high years. Low years are more like 20K, sometimes as low as 12K.
Comparing the relative heights of the Covid spike with the Flu "spikes" tells you something about the relative number of fatalities without needing to correctly account for any individual death.
2. Greater New York City:
I am going to start with the Diamond Princess data. Since everyone was tested, we have a fairly good dataset. Furthermore, the population density is about the same as Manhattan: 68000 per square mile. It's also the density of greater NYC, defined as the five boroughs and nearby cities in New Jersey, but not counting Long Island, Connecticut etc.
On the Diamond Princess, they had everyone interacting in close quarters until the first symptoms showed (about two weeks), and then they locked everything down. Not too different than NY. So we have a pretty good, but not perfect, proxy.
18% of the passengers and crew tested positive, and the fatality rate is 1.8%. However, cruise demographics skews elderly. The median age of cruise passengers is 65. The US as a whole has 14% of its population 65+. Since the disease affects primarily the elderly, there should be a correction of 14/50 to the 1.8%, making it 0.5%. That doesn't count the crew (none of whom died), so 0.7% is probably closer.
If you say 15M people in the NYC area, 2.7M of them will be infected if you treat it as a giant cruise ship. About 20K will die. Time will tell how accurate this model is. If you just take the NYC numbers and place them over the Hubei graph, extrapolate, and correct for the inner suburbs, you get 22K.
How many died of the flu in the same window? Assuming 34,000 deaths in the country, like last year, that would correspond to about 1600. A bad year would be twice that, and a good year half that. How many die in a year from all causes? About 130,000.
How does this extrapolate to the rest of the country? No idea. The 18% number will come down, but I have no idea how much. I'd need at least one more data point.