It will be more something like "sabbatical" in academia - you can devote a time to some side projects (whether career related or pure leisure), but everything you'll do you'll do with considerations of how your life will proceed once the sabbatical is over and there is no more safety net (assuming rational actors).
What happens after the money ends? They have to re-train, search for a job, apply, interview, explain a 5 years gap, climb the office/workshop ladder from the bottom again...
Situation changes radically if the income is guaranteed for life, if they are guaranteed to never have the necessity to work again.
Since a study limited to 5 years can't, by design, satisfy this last condition, it won't be significant at predicting how many people will leave a "hard" job.
Perhaps, when no longer totally living hand-to-mouth, during those 5 years they can e.g. take evening classes and eventually qualify for a better job.
Five years is much better but fundamental issue is still there - "there is a deadline after which the utopia ends".
Did you consider maybe using fewer participants but possibly much longer time-frames (e.g. for the same budget 10x less people but 50 years of BI, or even 20x less participants and 100 years of BI to have safe margin)?
I'm aware this would be hard / impossible to personally see to the full fruition (see e.g. Up series [1]), but even just that 5 years window frame for the observation and research (which you already prepared yourself for) but with participants who could have "piece of mind for the rest of their lives" could radically alter the results of the experiment.
------