The answer is no, for reasons that are hard to articulate succinctly. But the fact that the variance in "actual" expected value is lower than it initially appears makes "donating to learn" more attractive, and for this reason, among others, GiveWell recommended that donors split their donations between its top charities.