Interestingly, the CDC website no longer has these numbers. A week ago, their main COVID-19 landing page had a table of positive tests, negative tests, and pending tests (total tests was in the neighborhood of 450). That seems to have disappeared, with a subset of those statistics here: https://www.cdc.gov/coronavirus/2019-ncov/cases-in-us.html.

I've heard this argument but it skews in both directions, I imagine. You'll have mild cases that recover where no test is ran that skews the data to the "more harmful than being presented" but I suspect there are also cases, especially for those with other preexisting conditions, where deaths also didn't run a test. Those cases would skew the data in the opposite, "less harmful than being presented" side. You essentially have a huge mess of amalgamated/non-systematically collected data. Even in controlled settings you get a lot of censored data to deal with.

RT-PCR tests are reasonable these days in cost, but it isn't cheap by any means and I suspect many early cases weren't checking or didn't have the capability to check. The question is: which way is the data skewed more towards?

I know absolutely nothing about the chinese healthcare system. Even if you do, there's a significant amount of guesswork. Only now as it spreads are we starting to get more controlled tests that give us more accurate data.

For now, I have a tendency to trust the raw data until I see definitive evidence of skewness otherwise.