Because in South Korea the rate is more like 0.5% and they've done the most testing. As far as I can tell, basically all the data around this epidemic is so du

Because in South Korea the rate is more like 0.5% and they've done the most testing. As far as I can tell, basically all the data around this epidemic is so dubious it's nearly useless. We use it anyway because ignoring data is really hard when the alternative is intellectual darkness. Yes, that's a strong claim, but look at the error bars! Above in this thread we have someone giving two estimates for total dead in the USA that vary by a whole order of magnitude. A difficulty with all current rates is that nobody knows the denominator: how many cases are there really? Some cases are asymptomatic, many are mild, people got the message by now to stay home and try to avoid overloading the health system so they aren't necessarily going to turn up to get tested unless they start to feel really quite bad. To know this you'd need mass random testing of the population, but nobody is doing that, not even in South Korea. If you divide deaths by cases that turned up at hospital you get these higher rates. But that can't then be multiplied by population figures to derive total expected death rates. It's more like a "death rate of those who make themselves known because they have severe symptoms". Another problem with these stats is people tend to interpret them as "number of people who will die who otherwise wouldn't have", i.e. imagining a drop in the population graph. But many of those who are dying were about to die of old age or other factors, so deviation from standard death rate isn't going to be that high. The causes of death will shift, but the total number will be less than a simple "Chinese rate * population" calculation.