Debugging spreadsheets is pretty notoriously difficult for showing that it does the expected thing, let alone figuring out what assumptions have been made and how valid they are

I think the interesting point is that apparently very few people even tried. I don't know how everyone is dancing around that in this thread but that's what Caplan seems fascinated by in this blog post. He's not saying he's right because nobody checked them, or that everyone ought to have. He's fascinated that nobody chose to challenge his conclusions on quantitative grounds. And he gives some anecdotes that suggest that this is because people in the field are not interested in crafting optimal policy -- which does almost by definition require some amount quantitative approaches -- they are interested in publishing.

I think he actually believes the math. The person who would do it as a troll, as a reductio ad absurdum against the inputs, is me. Except I'm too lazy.