Earlier quoted context omitted.
By lack of a pile of bodies.
If there’s a lot of uncounted asymptotic cases the IFR is correspondingly lower.
The much-higher, 50% numbers that were being thrown around were obviously overestimated - precisely because it takes a week or two or three for some people to start showing symptoms.
As such, it's incredibly unlikely that the # of real cases is some large multiplier of the number of confirmed cases in any areas that are regularly testing people with flu-like symptoms.