The problem with this study and many others that report on TDD's benefits is that they don't compare the results with a test group that writes unit tests after

The problem with this study and many others that report on TDD's benefits is that they don't compare the results with a test group that writes unit tests after the fact. So, it's not at all clear what they're validating: the benefits of unit tests (which I think are well accepted) or the benefits of writing tests before code. I suspect it's primarily the former.