According to David Paul Kuhn, "the most durable reality of American politics" is that "white men matter most." Dana Goldstein offers an excellent substantive rejoinder, but I'm always perplexed by this broader genre of analysis. The presumption often seems to be that Americans vote via some kind of demographic electoral college and thus we can argue about whether "white men" or "Hispanics" or "married women" or "independents" are the key to the election in much the same way that we can identify some states (Ohio, Florida) as big swing states, while others are either less important because small (New Hampshire, New Mexico) or else unimportant because non-swingy (New York, Texas).
Obviously, though, elections don't work like that. I imagine that only a very small number of black lesbians voted for Bush in 2004, but convincing one of them to vote Democratic in 2008 is exactly as valuable as it would be to convince any other Bush supporter -- white, green, red, blue, whatever -- to flip sides. The only relevant distinctions for these purposes have to do with which state someone lives in.
It is true, of course, that since most people are white people, white people "matter most" in some crude numerical sense. But men don't matter more than women on this score. And the bigger you build up your demographic group the more its "importance" comes at the expense of any kind of precision. Given that white men are something like 40 percent of the population, it's too big a group to target in a meaningful way.