[Hupriene]

I agree that the numbers in the US are unlikely to be changed post-collection, but there are still pressures that lead to the under reporting of suicide rates here as well. For example, accidental overdose can be very difficult to distinguish from suicide. If there is any question, coroners will often err on the side of ruling the death an accident to spare the survivors additional emotional hardship.

[Pharmakon]

A good point, but it means you can make a high estimate by including a statistical model accounting for a percentage of accidental deaths (including overdose) that are likely suicides. As long as you have all of the data, and it’s reliable, the rest is easy. Misclassification is a much easier issue to deal with than outright fiction. The problem with China is you can’t tell what if any parts of the dataset are reliable so no methods can be employed short of espionage.