> do not show that there are negative and positive contributors, only isolated incidents any person can choose between, and whether or not this sums to a positive or negative person in relation to eg. their tax contributions

How is "there are negative and positive contributors" different from "whether or not this [behavior] sums to a positive or negative person"? There's no reason Simpson's paradox would apply here; and, Simpson's paradox with respect to what? We are trying to find if a specific person X is a negative or positive contributor to local QoL.

Your example actually illustrates that and contradicts the first point - if we introduce a free service, let's say QoL would go up; someone moves in to abuse it, you admit QoL goes up less ("smaller increase than they hoped for") - that person is a negative contributor, all other things being equal. Or, if you say their net is still positive (no reason it would always be positive), then the original complaint - states punished for improving QoL - does not apply. They got rewarded by this positive net contribution, not punished, right?

As for the example itself, there's nothing "free" - someone pays the doctors providing healthcare, etc., and you could argue the payers' QoL goes down. If you don't limit how much people may have to pay (it's universal vs a limited number of taxpayers), their QoL can go down a lot. If it doesn't balance out and the state gets "punished", it's precisely because it's bad policy that doesn't improve QoL. The debatability of the merits of universal healthcare itself aside, the reason it may not be too bad on country level is because it's hard to move in to make use of it; it comes with status/residency restrictions, and also health-based immigration restrictions in some countries (like Canada, IIRC).

I do admit I generally think states can improve QoL much less than they think they can, and can degrade QoL much more than they think they will, but there are many things states can do (and have done in the past) that create good incentives and don't suffer from this problem. Most of them actually involve undoing things. Others involve things that incentivize more positive contributors than negative as 2nd order effect; be it safety, building certain kinds of infrastructure, etc.