I'll point out that "saving lives" is actually a pretty weird unit, and the more seriously people start to take it the more important that weirdness gets.
For example, suppose there are a thousand people who suffer from Invented Syndrome, a congenital condition that causes periodic life-threatening pulmonary arrests. Suppose I have two treatments available: treatment A ends the pulmonary arrest, saving the person's life, but does not prevent subsequent arrests. Treatment B not only ends the arrest, but also cures the condition, ensuring that no subsequent arrests occur.
If I'm counting saved lives, it seems I end up favoring treatment A, since each use of it saves a life... I predict more lifesaving events if I manufacture and distribute A than B.
Which, you know, is fine... if that's what I value, that's what I value. But I suspect most people who claim to value saving lives wouldn't actually go for that. (Nor would they go for the middleman-eliminating version where I put a thousand people in a warehouse and forego pressing the button that kills them every few seconds. Heck, I can sell you saved lives at much much cheaper than $800 a head at that point.)
And while they might not admit it, I bet they'd also prefer to save an otherwise healthy seven-year-old with an expected future lifespan of 70 years than an otherwise healthy eighty-year-old with an expected future lifespan of 7 years.
Perhaps what they actually value is human observer-moments?
Given some plausible assumptions, increasing the probability of a future trans-humanist utopia populated by an unimaginably large number of humans should be much higher return in terms of observer moments:
From Yvain's 'proposal' to measure money in dead children:
This makes sense to me, to a limited extent. You can spend money for your own benefit or to help others elsewhere, and there really are people who wouldn't have to die if you would forgo some luxuries. Making this tradeoff more explicit ("we're looking for an apartment costing no more than six dead children annually") might lead some people to greater generosity. It's a way of abstracting compassion.
Two things worry me, though. The first is that there's a big focus on spending here [2], but increasing earnings deserves more focus: getting a raise or a new job that added $10K to my salary would let me keep more children from dying than would reducing my spending on myself to zero. [3] The second is that thinking of all your purchases in terms of dead children is likely to make you miserable. Not just that, but miserable to little gain: you still probably spend almost as much money on yourself, you just feel more guilty about it. Much better, I think, is to pick a rule for how much to give and then apply it to money as it comes in. That way each purchase has no effect on the number of deaths you're averting.
(Note: I also posted this on my blog)
[1] The current number is probably closer to $2K.
[2] Maybe this is because it sounds weird to talk about salary in terms of dead children? ("I wonder what job earns me the most dead children?") Perhaps for earning the unit should be the "undead child"?
[3] In 2011 Julia and I lived on $18K for the two of us, not including taxes or health insurance.