I take it that you're suggesting marginal analysis based on the standard correct classical causal decision theory (in which no one is responsible for saving a life by donating blood unless someone would have actually died had that donation not been made) out of either belated humility about the probability of an SIAI-originating decision theory being correct, or because you're planning to actually convince someone and you don't want to invoke Hofstadterian superrationality in place of the standard correct decision theory?

:)

My guess would be that at the margin, a blood donation saves less than 0.00001 lives. (Otherwise, compensation would be increased for the paid donors). But, if you want to use a TDT/UDT style analysis, here are some relevant statistics from the American Red Cross:

• The number of blood donations collected in the U.S. in a year: 16 million (2006).
• The number of patients who receive blood in the U.S. in a year: 5 million (2006).

Given these numbers, I would estimate that roughly 0.5 million (US) lives are saved (more accurately, extended) by blood products annually. If you adopt the assumption that all blood comes from voluntary, uncompensated donations, and divide those 0.5 million lives among the 16 million annual donations, you get one life saved for every 32 pints donated - not as much as jsteinhardt hoped, but still significant enough to earn a major warm-and-fuzzy.