I think the idea is more that ineffective (or at least terribly inefficient) programs would be much less likely to survive in such an environment. If program a is designed to achieve effect B, some people will argue for it because they think that B is desirable, and others will argue against it because they think that B is undesirable, or because it produces some side effect C which they find undesirable. However, if it is shown that program a does not achieve effect B, or is terribly inefficient at doing so, then even proponents of effect B would have to agree that it should be eliminated. Presumably, those proponents would then push for some new program aimed at achieving effect B, but that's a separate issue.