I find it hard to take this sort of analysis seriously, when the authors admit that they don't even understand the order of magnitude of the error bars on the data.

Not to mention that the regulatory landscape has changed over time, and since that directly influences several classes of errors (from waiters under-reporting tips to "the 1%" hiding money in now-illegal tax shelters) estimating historical trends is an exercise in futility.

And we effectively have no way to benchmark how well these different studies are doing at measuring the underlying phenomena that they're trying to measure, this is really just making up numbers and calling it "wealth distribution".