[InsideOutSanta]

There will never be proper studies with control groups to test exactly how harmful beating children is, so this is an unrealistic standard to expect. Given this context, the person you're responding to is correct: we have overwhelming evidence that corporal punishment is harmful in general and very harmful for children.

[dash2]

The point isn't just that we can do RCTs. The point is that the methods used are not even adequate on their own terms. Just as one example, the standard method with longitudinal data would be to throw in individual fixed effects. But they don't do that. Another example: I know of no serious cross-country panel analysis with (say) time and country fixed effects to examine the effect of national spanking bans. There is a cross-country cross-sectional analysis, which is just not adequate to draw any conclusions.

Even if the methods were the best possible given the difficulties, you wouldn't then say this was "overwhelming" evidence. You'd say "the best evidence we've got" and you'd then assume that parents don't know nothing and exercise a bit of humility. (Though to be fair, that argument does not generalise to the Singapore decision-making authorities! Maybe they don't have any deep local knowledge that should lead us to trust their judgment.)