If some will lead to the survival of babies like Olivia (which is not, considered in itself, identity-affecting) but also to the non-conception of (possible future) babies like Jack (an existential effect). But at least we can say of the policies in the first group that, while they may indirectly affect who comes to exist, this is not their main aim. The actions and policies in the second group, conversely, are directly identity-affecting and altering the constitution of the future population is part of their aim (although those involved may not conceptualize it in quite these terms).
These policies determine which (possible future) people will come to exist. For example, if we successfully encourage prospective teenage mothers to delay their pregnancies by ten years then we are getting them to have different children from the ones they would otherwise have had (and/ or to have fewer children). When, following PGD, we choose between embryos for possible implantation we are choosing between one (possible future) child and others. When we separate X-chromosome sperm from Y-chromosome sperm and choose to use one or the other for sex selection, we are choosing between different (possible future) children and may, rather obviously, end up with a boy instead of a girl.
And the same goes for a woman who is inseminated by MENSA members, rather than her husband: she is choosing to have a different (possible future) child. Why do we generally regard such policies as identity-affecting? It may be because we accept what Parfit calls the Time-Dependence Claim.