never

If the question is about a program that has already been or will be implements, then you would want to use a pre-post test -- that is, you want to devise an experiment or quasi-experiment that can be used to assess whether the policy had an effect. These kinds of models are defined/discussed by in the well-known work of Campbell and Stanley (1963), Experimental and Quasi-Experimental Designs for Research.

If the policy fails, then obviously its not worth a monetary investment. But, even if the policy works, it might not be worth the expense. So, if the policy works, THEN you can do a cost-efficiency analysis. For example, lets say that you randomly assign people to an experimental (E) and control (C) group as in a classic experimental design. The E group gets medicine A, and the C group doesn't get the medicine. You find out that the medicine has a significant effect, and that those in group E have a 50% less chance of getting the disease. But, lets assume the medicine costs $ 1000 dose, and you have to take 30 doses -- or you have to spend $30,000 and have some chance that you don't get the disease. Then, you need to have some calculation of the cost of the disease, and whether the $30,000 you are going to spend for each person is worthwhile. Do the people who get the disease lose work days? How many? What is the cost of the lost work days? Maybe the disease isn't serious (e.g., its the common cold), so suppressing it with $30,000 of medicine isn't cost effective.

In any event, the question you raise require a multilevel analysis -- one examines the effectiveness of the program, and the other examines the cost of the program relative to its effect.

thank you very much for the answer and time for me and of course very helpful ... by looking at the long answers that have been explained in detail, the analysis needed is success and failure after the implementation of the policy program

@Heather Douglas

@Michael J. Lynch

Cost-benefit analysis (or benefit-cost analysis) is a type of policy research, a project that is ususall done by quantifying all the facotrs involved and monetizing them, thus a qualitative study of the policy would e a separate, although related project. In other words, a CB analysis cannot be a qualititae research.