If you experiment properly, you take people who were in a region on $startdate and give them UBI indefinitely, also levying enough additional tax or welfare cut

If you experiment properly, you take people who were in a region on $startdate and give them UBI indefinitely, also levying enough additional tax or welfare cuts on that region to pay for it. That's geographically more limited, and it might be temporally more limited if it turns out to be a massive failure forcing the curtailment of the whole thing, but it's very different from testing to draw the amazing conclusion that penniless people given $7.5k are on average more financially secure a year later, which actually tells you less about a hypothetical UBI than the existing welfare state. (There might be other public policy reasons to test giving cash handouts to homeless people, like "is this more efficient use of funds than spending the equivalent on homeless hostels and counselling?" but likewise, you have to construct the experiment to actually look at that side of things)