None of these empirical studies are useful for broad policy decisions. This is a bit like linear regression, there are to many dimensions each matters different

None of these empirical studies are useful for broad policy decisions. This is a bit like linear regression, there are to many dimensions each matters differently and the same classifier will not hold true if you add more dimensions randomly or remove few. There is no "core human trend" to all this that will get revealed only by empirical studies but not by logic. (Empirical studies are useful to judge say whether vaping actually makes us more healthy on average. Smoking is so irrational at some level that no amount of logic can help us predict how people behave.) The economics logic is pretty clear. People will react to incentives. But there is no hard formula that we can apply. Raising minimum wage by few cents will not cause job loss but increasing it by few dollars will. But if you make minimum wage around $100 then whole industry might go underground and no effective jobless but for in government record keeping. I think effectiveness of UBI will be useful only if we perform far too many experience under far too variable circumstances and then understand the broader trends.