Choice of comparisons of anti-poverty approaches (or many policy analyses) are driven by the list of the top most politically probable.
For example, you don’t compare 9, 10 and 11% options in modelling the effects of a shock or government policy. You model x% size change of the most likely options.
Which yes, not not imply optimality. But unless you’ve got 20 years to iron out kinks, etc., while publishing a dozen articles on the way, probably the more re
So you model the effects of a politically feasible sized option for 2 or 3 different approaches. Rather than optimizing for some particular varirable or another (generally utility and/or production/consumption), then, you’re comparing the general direction and magnitude of some different option.
Because often debates are not on margins, but about completely different approaches to addressing an issue that has been prioritized for being addressed.
So, if you want to improve child nutrition, the focus is rarely on whether a $5, $10 or $20 a month subsidy for school feeding programs is more optimal (although that might be of some interest to decision makers). Rather, the question can be comparisons of politically feasible different approaches. For example, then, to compare the politically feasible $5 a month subsidy for school feeding, the politically feasible cash transfer to the family of $10 a month if school attendance is proven, or the politically feasible.
And then, among the politically feasible options, to outline some costs and benefits (maybe without overly much concern to put a decimal place on “optimality” according to some narrow criteria), in particular which of the options is more beneficial to the narrow target, and which are more likely to have positive contributions to other targets (e.g., primary graduation rate).