Bradley writes (p.300):
Remember that death is a harm that is merely extrinsically bad for its victim. Death is harmful because of what it prevents; it has a sort of extrinsic value. When determining whether there is any reason at all not to perform the action, we should never consider the extrinsic values of the results of the action. For example, suppose you save my life by curing me of a debilitating disease and, as a result, I live an extra ten very happy years. I then die from an unrelated cause, and my later death deprives me of yet another ten happy years. Your saving my life was an indirect cause of my later death, and my later death was very bad for me. But my later death was merely extrinsically bad, and in evaluating how good it was that you saved my life, the extrinsic badness of my later death is irrelevant. Before my death I would not be justified in complaining, “Yes, you saved my life and gave me ten happy years, but you also indirectly caused an event that will prevent me from getting another ten happy years; so the bad you caused cancels the good you caused.” The reason this would be an unjustified complaint is that your saving my life, rather than letting me die, did not prevent me from getting those later ten years of a happy life that my later death prevents me from having. Similarly, in the case at issue in McMahan’s argument, the act of allowing a person to come into existence, rather than preventing his existence, does not prevent that person from getting the goods of which his death deprives him. Even though the act of allowing the person to come into existence indirectly causes another event that is very harmful to the person, that fact is not relevant to the evaluation of the act.
The upshot of this argument is that it matters greatly how we would prevent the harm of death. We have very strong reasons to prevent death by ensuring the continued living of the subject. But we have no such reasons to prevent death by preventing the subject from ever existing at all. The harm of death, on Bradley's view, is just that it deprives the subject of future goods, but non-existence is no better in that regard.
Does that really seem right, though? I'm dubious. Consider two possibilities in turn. First, suppose we accept some broadly "biological" view of identity, such that an embryo is one and the same entity as the person it grows into. Here we run into the spontaneous abortion objection: it just doesn't seem true that we typically have very strong reasons to save the lives of embryos that would otherwise be spontaneously aborted (with no-one aware that a pregnancy had even taken place). There's no real "harm" there.
On the other hand, suppose we instead accept a broadly "psychological" view of identity, such that we don't come into existence until we acquire something close to mental "personhood" -- as toddlers, say. It's now clear that we have reasons to ensure the continued living of the subject, but the second claim is now in doubt: should we really think that the death of a toddler is no worse than simply failing to conceive it would have been? The early death of a person seems to me positively bad in a way that mere non-existence is not. (Do you agree? It's obviously instrumentally worse, e.g. for the emotionally-invested parents, but even bracketing all that... it just seems worse, in itself, to have more people exist with brief, tragic lives.)
This might be taken to suggest that death is not only extrinsically bad after all. Early death, on this alternative proposal, introduces an element of tragedy that plausibly contributes positive disvalue to a state of affairs. So even if Bradley is right that the extrinsic badness of early death is no reason to prefer non-existence, this additional aspect of death's badness might furnish such reasons after all.
Another possible explanation is opened up if we accept actualist partiality. On this view, it is only for people who actually come into existence that we can be presented with "personal" (as opposed to merely "impersonal") reasons. So, in cases of early death, we have personal reasons for regret that we would not have in the case of mere non-existence. (There we could at most have impersonal reasons for regretting that the world doesn't have an additional happy person in it.) We would have additional reasons for thinking that things had turned out badly -- that it would be so much better had the person lived on. Had they never existed, we would not have such strong reasons to prefer that they had existed.
But this does not yet get us the conclusion that early death is worse than non-existence. To get that, it seems we need some further principle to the effect that it's better to prevent deeply regrettable events. But this brings us back to Bradley's original argument (just with "deeply regrettable events" substituted for "great harms"). So I don't think actualist partiality can do any real work here. If early death is worse than non-existence, then this must simply be a basic datum in our (impersonal) axiology.
What do you think?