I'd agree with that, because refactoring is hard, especially if you don't intimately understand the domain, and I think "effective government" is much less understood than "high quality software" as a whole, so it's extremely hard to refactor government. Additionally, there are more politics involved in gov't refactoring than code refactoring, so even provably good ideas can get shot down. All the same, I endorse "learn more about the problems and get better at refactoring" over "rip things out" virtually always.