That would be Italy. Which is very worrying.
Not really. Italy is not doing much testing, and the testing distribution for age does not reflect the age distribution. Their fatality data is a severe overestimate, which is why it is so high.
Bear with me:
Assumption: No country is cheating by underreporting deaths. So, if someone dies from corona then it counts as death from the corona.
From this assumption, I would say that countries with the lowest fatality rate (and a high number of testings) are closest to the real fatality rate. This is simply logical if every death fro corona is reported as a death from corona, then under no circumstances, the official fatality rate can be higher than the real fatality rate.
Assumption 2: Every country misses some people who are infected (even Germany and South Korea).
This means that the official fatality rate is again an overestimate of the real fatality rate. Essentially, the 0.2% fatality rate in Germany is an overestimate. Same for South Korea's 0.7%. Or Bahrain's 0.3%.
Now, it is also not a coincidence that there is a negative correlation between the number of testings and the number of deaths. The more testing, the less deaths (Singapore, Korea, Germany, Bahrain). This strengthens my assumptions. Those countries are having fewer deaths simply cause they are finding more young people who do not show symptoms as positive for the virus. By simply testing more people.
I think that the logic is relatively sound, though there are some holes in it:
1) My assumptions do not count for age. But it is not a big problem, we already know that most of tested people are oldish (cause they get sick easily, and then do a test). Again, any fatality rate is overestimate. Also, not many kids are being tested. With kids surviving but not being counted as infected, again any official fatality rate is an overestimate.
2) My assumptions do not count for different strains of the virus (a possibility, but still we do not know much except there seems to be two different strains of it).
3) Some people who are sick will die, and so the fatality rate will increase. This is the only point that can increase the official mortality rate. But then, Germany has only 2 critical cases, and Bahrain has 3.
4) There are 2 viruses which are similar to each other, and analysis do not count for it. I think this is a very wild guess, with extremely low probability.
If I have to bet, I would say that the true fatality rate is somewhere between 0.1-0.5% assuming that the medical system is working and the assumption 1 is correct (assumption 2 is correct).
Again, this is hardly a scientific analysis, and for that much more data is needed. It is just me finding it hard to concentrate on my real work. But I think that the logic does not have many bugs.