I'm predicting an extended squabble about this, which is probably gonna get technical to the point where I won't have an informed opinion. We like to think of data collection as a sort of immaculate conception of information, but it's a really involved process, moreso if it's not a category that's already tracked monthly/quarterly by a big federal agency.
I remember the kerfuffle over "medical bankruptcies" during the bankruptcy reform bill's passage, where Elizabeth Warren and others argued that most people go bankrupt because they're hit by unexpected medical costs. This got "debunked" from the right, showing that they used a very broad definition to say medical expenses "caused" people's financial problems, but then replaced it with a "real" figure that was largely the result of tilting the assumptions the other way (stuff like not counting prescription drugs as a medical expense).
Galbraith's critical piece in Dissent called out Piketty for overselling the value of tax records as the only good source, but he said that Piketty's results were in line with what he'd seen from payroll data. I don't have a reason to particularly trust Pikkety, but the phrase "simplified and cleaned up the data" doesn't immediately fill me with confidence either.