When you really peel back the layers, you'll find that voters' instincts are that any means-tested program should come with a hefty punishment for using it. While they aren't exactly against helping, they definitely think the priorities are 1) Spending as little taxpayer dollars as possible, 2) Punishing any recipient of help enough to be a warning to others, and then, distantly, 3) Helping.
This is not exactly voters. What really happens is more like this: There is some call for helping people, voters want it, but voters (possibly some of the same voters, possibly different ones) also want lower taxes.
So the plan that passes is the one that claims to help people (appeasing the first group) while also costing the minimum amount of money (appeasing the second group). The proposal with evolutionary fitness in politics is then to create a system that helps people on paper but actually limits eligibility for the system to a minimum number of people or otherwise makes use of it more arduous to deter usage so it costs less. Extra points for making it so complicated that people don't realize what's really happening so they don't object if it doesn't really do what they wanted.
But there is also a level of incompetence/inefficiency here, because the complex and overlapping systems that pass as a result often have high overhead that waste tax dollars, or create bad incentives because nobody thought them through and those incentives create deadweight economic losses adverse to the interests of even the people who want to minimize taxes. As the obvious example, if you create a system with welfare cliffs and then people have the incentive to stay on welfare instead of taking a job, you now have more government expense and lower tax revenue than a system that doesn't do this.
At which point there is a Pareto-optimal improvement on the table if you can get the bill passed.
