There are a lot of things that make calculating a straightforward percentage hard:

- not all cases detected

- deaths are a laggy indicator

- recoveries are an extremely laggy indicator

- impact of demographics, healthcare quality

- sizes of effects of all 4 previous things depends on the country and the time

This has led to internet commentators coming up with wild numbers, ranging from 0.1% to 20%. At the risk of sounding like a spoilsport, I suggest we all defer to the expert numbers (2-3%), because the experts are perfectly aware of all 5 effects I mentioned, and more, and have already tried their best to correct for them, using better data than we have.