It seems like we’re in a Catch 22 scenario. Models and experiments only give so much information by the very nature of models and experiments simplifying much more complex problems, and in order to collect the kind of holistic data that would speak to a societal level would require an experiment that is functionally identical to just full implementation. Like experiments can only get so big or go on so long before it just becomes the actual thing itself and is no longer an experiment.
And in regards to your other comment about dialing in from the extremes, instead of making everything a uniform number, we can use an formula and a handful of variables to arrive at a local number. A function that takes in common cost of living costs, such as food, shelter, clothing, utilities, transport, and generates a number that will adequate cover those expenses for a given area plus some extra because something unexpect can always happen.