> Isn’t a cliff simply a ramp with infinite slope? I would suspect the problem is that if we can’t tune something as simple as a step function correctly (which is just an x intercept), it seems even less likely we would be able to tune a ramp (more variables - x intercept and a slope, and that’s assuming it should even be linear instead of e.g. quadratic).

One of the challenges of tuning means-tested welfare phase-outs is that means-tested welfare consists of dozens of different programs, most of which have additional qualifications besides the means test, governed by different levels of government, often with different definitions of what income counts against the phaseout.

UBI makes that simpler by consolidation.

Something as simple as “for each $1.00 you earn above the eligibility threshold, this benefit will decrease by $0.50.” would do the trick.

For those close to the line, there would always be a benefit to accepting that slightly better job.