The only way failing to save lives can be equated with killing people is by subscribing to pure utilitarianism. But by that philosophy, contraception is also equivalent to killing people: the end result is that fewer people are alive than in the counterfactual case where you had children. The counterargument that contraception is not immoral because you aren't obliged to have children is fine, but it also applies to the other case: you aren't obliged to give your money away either. In other words, we don't actually subscribe to pure utilitarianism, so we shouldn't pretend we do.

That having been said, my guess as to the best way to help people in the Third World -- the way with the highest expected utility, as well as the most assurance of positive rather than negative utility once all the consequences have worked themselves out -- is to support economic development in poor countries, to help them get to the point where they aren't poor countries anymore. In other words, to buy goods and services produced in the Third World.

Counterintuitive? Perhaps, but remember what our intuition evolved for: the gain of social status by giving donations, thereby establishing ourselves as having higher status than the ones to whom we give. To engage in a mutually beneficial transaction is to acknowledge the other party as our equal; no wonder it doesn't trigger the altruism instincts. So if our goal is to actually do the most good, we should expect the method to be counterintuitive.