A new paper by Daniel Kruger posits a general framework for why women tend to live longer than males across cultures when environment is equalized (obviously there are cultures where strong sex bias in quality of life is more important than underlying genotypic variation). The idea is simple, it is a spin off of antagonistic pleiotropy where males tend to engage in risky behavior when young and so increase their mortality in the interests of increasing their potential reproductive value. Additionally, for a host of of genetic and physiological reasons we are also susceptible to many diseases which women aren't. The combination of endogenous male vulnerability to a variety of diseases along with particular behavioral tendencies results in elevated mortality. The male strategy of engaging in risky behavior when young makes some evolutionary sense, as male sperm is cheap and plentiful most species seem to show greater male reproductive variance than female variance. Extreme cases like elephant seals show us the reductio ad absurdum. Interestingly, Kruger points out that low SES males tend to show a greater mortality gap with equivalent SES females.
Addendum: In The Mating Mind Geoffrey Miller posits that male reproductive skew, and the correlation of this skew with genetic fitness, allows our species to expunge its genetic load.1 - Elevated testosterone tends to weaken the immune system, and sex linked traits.2 - Imagine a locus which controls the level of "buck wildness" of an individual. Let's imagine that the variation of the expression of the locus is modulated by the basal testosterone level in the case of polymorphisms. This makes it so that females won't really exhibit a big difference because of genetics, but males will. Imagine a "stay cool" allele and a "buck wild" allele. SC & BW respectively. Assume additivity, so that in terms of "wildness" SCSC = 0, SCBW = 1 and BWBW = 2. Now, assume that mortality at age 15 is linearly proportional to "wildness." That is, the more wild you are the more likely you are to get killed. But, let's also assume that fecundity is non-linearly proportional to "wildness" so that there is an exponential tendency to have more offspring in relation to wildness. In this case, it is clear that wildness is worth the risk of death, because the locus is being passed around in many individuals and elevated morality is more than made up for the fact that the allele is carried by individuals that won the "love lottery." Or, one can imagine a fecundity distribution, the wilder males would have a larger area under the curve (the integral) than the calmer males, even if you noted that many more wild males tended to have "0" offspring because of mortality.